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 95F703858C78; Wed,  7 Feb 2024 08:04:30 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 95F703858C78
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1707293070;
	bh=4jXoZOkDZ5Xj1mT3Uvyd0XCT9sSboUW/CQ5FL8nTcUc=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=m4GW4z78oNJCrm21jWJJwbMY/XU+rMomY90uba5bJ4qJlDL4XM6Iv8bXYRUU1LZ5I
	 mpqpBY/aFyihFdf557bhIAMfVRP77AD1nliz/sX+lm216JElhnYrq3Ad544LCl5ZqP
	 mk8kPF7j5QO/KhnPOeAYyEFevsa16/aaU/Za2Tqg=
From: "rguenther at suse dot de" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/113583] Main loop in 519.lbm not vectorized.
Date: Wed, 07 Feb 2024 08:04:29 +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: missed-optimization
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: rguenther at suse dot de
X-Bugzilla-Status: UNCONFIRMED
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-113583-4-fM4UnFXTwg@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-113583-4@http.gcc.gnu.org/bugzilla/>
References: <bug-113583-4@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=3D113583

--- Comment #14 from rguenther at suse dot de <rguenther at suse dot de> ---
On Wed, 7 Feb 2024, juzhe.zhong at rivai dot ai wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D113583
>=20
> --- Comment #13 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
> Ok. I found the optimized tree:
>=20
>=20
>   _5 =3D 3.33333333333333314829616256247390992939472198486328125e-1 - _4;
>   _8 =3D .FMA (_5, 1.229999999999999982236431605997495353221893310546875e=
-1, _4);
>=20
> Let CST0 =3D 3.33333333333333314829616256247390992939472198486328125e-1,
> CST1 =3D 1.229999999999999982236431605997495353221893310546875e-1
>=20
> The expression is equivalent to the following:
>=20
> _5 =3D CST0 - _4;
> _8 =3D _5 * CST1 + 4;
>=20
> That is:
>=20
> _8 =3D (CST0 - _4) * CST1 + 4;
>=20
> So, We should be able to re-associate it like Clang:
>=20
> _8 =3D CST0 * CST1 - _4 * CST1 + 4; ---> _8 =3D CST0 * CST1 + _4 * (1 - C=
ST1);
>=20
> Since both CST0 * CST1 and 1 - CST1 can be pre-computed during compilation
> time.
>=20
> Let say CST2 =3D CST0 * CST1, CST3 =3D 1 - CST1, then we can re-associate=
 as Clang:
>=20
> _8 =3D FMA (_4, CST3, CST2).
>=20
> Any suggestions for this re-association ?  Is match.pd the right place to=
 do it
> ?

You need to look at the IL before we do .FMA forming, specifically=20
before/after the late reassoc pass.  There pass applying match.pd
patterns everywhere is forwprop.

I also wonder which compilation flags you are using (note clang
has different defaults for example for -ftrapping-math)=