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 E75703858D1E; Fri,  5 Jan 2024 21:03:59 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org E75703858D1E
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1704488639;
	bh=n/5h/QSHOSMt6t3auwxMJDRZwI6mGNY/ry8IOIqFJH8=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=n9o/pJ+9YzMVVEDsAjmUT7HGBz+pPY0a12bnz4z0SxwOg+gJgB5b4PsvR+hgWGCwr
	 5cGnoT0u92nb3yIoXYC03eqwVF2gpxPALK5Eitpz6tbCpR/SJ9dCcmIEQad47ejcZ1
	 rTLldyGjHTzau7oAEeoDrtBAbbmx9gFXBl9xrECA=
From: "hubicka at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/113235] SMHasher SHA3-256 benchmark is almost 40% slower
 vs. Clang (not enough complete loop peeling)
Date: Fri, 05 Jan 2024 21:03:59 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: target
X-Bugzilla-Version: 14.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: hubicka at gcc dot gnu.org
X-Bugzilla-Status: NEW
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-113235-4-wKCU19smXd@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-113235-4@http.gcc.gnu.org/bugzilla/>
References: <bug-113235-4@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=3D113235

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
The internal loops are:

static const unsigned keccakf_rotc[24] =3D {
   1, 3, 6, 10, 15, 21, 28, 36, 45, 55, 2, 14, 27, 41, 56, 8, 25, 43, 62, 1=
8,
39, 61, 20, 44
};=20

static const unsigned keccakf_piln[24] =3D {
   10, 7, 11, 17, 18, 3, 5, 16, 8, 21, 24, 4, 15, 23, 19, 13, 12, 2, 20, 14,
22, 9, 6, 1
};

static void keccakf(ulong64 s[25])
{=20=20
   int i, j, round;
   ulong64 t, bc[5];

   for(round =3D 0; round < SHA3_KECCAK_ROUNDS; round++) {
      /* Theta */
      for(i =3D 0; i < 5; i++)
         bc[i] =3D s[i] ^ s[i + 5] ^ s[i + 10] ^ s[i + 15] ^ s[i + 20];

      for(i =3D 0; i < 5; i++) {=20
         t =3D bc[(i + 4) % 5] ^ ROL64(bc[(i + 1) % 5], 1);
         for(j =3D 0; j < 25; j +=3D 5)
            s[j + i] ^=3D t;
      }
      /* Rho Pi */
      t =3D s[1];
      for(i =3D 0; i < 24; i++) {
         j =3D keccakf_piln[i];
         bc[0] =3D s[j];
         s[j] =3D ROL64(t, keccakf_rotc[i]);
         t =3D bc[0];
      }
      /* Chi */
      for(j =3D 0; j < 25; j +=3D 5) {
         for(i =3D 0; i < 5; i++)
            bc[i] =3D s[j + i];
         for(i =3D 0; i < 5; i++)
            s[j + i] ^=3D (~bc[(i + 1) % 5]) & bc[(i + 2) % 5];
      }
      s[0] ^=3D keccakf_rndc[round];
   }
}

I suppose with complete unrolling this will propagate, partly stay in regis=
ters
and fold. I think increasing the default limits, especially -O3 may make se=
nse.
Value of 16 is there for very long time (I think since the initial
implementation).=