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 067273858CDB; Sat,  9 Dec 2023 07:48:02 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 067273858CDB
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1702108082;
	bh=/OyEHZ4oPap3AqRZyxcoUw5pMYQCYVwJxLOXNFNVQQM=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=Tf8U5/6IrqHNh1qdnJFPbRjKw0zElJjYZwrdBO/UewGEL8TSYTnDjzDIE5ZTf97B+
	 HYVKtyQFS9ZVPoYgpGwgz3/yYIv+mpe1+jlBoLUZf54nBLg1EKcdLhYLLlPnqcKnAl
	 QopyVovgGqO/a0PtFLUy5WlakNyLuSXh3ta6k/PE=
From: "xry111 at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/112935] [14 Regression] Performance regression in
 Coremarks crcu8 function
Date: Sat, 09 Dec 2023 07:48:01 +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: xry111 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: 14.0
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-112935-4-0sqpQ3Ddo1@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-112935-4@http.gcc.gnu.org/bugzilla/>
References: <bug-112935-4@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=3D112935

--- Comment #8 from Xi Ruoyao <xry111 at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #7)
> (In reply to Xi Ruoyao from comment #5)
> >=20
> > so we still slightly penalty multiplication.  To me we should code
> > COSTS_N_INSNS (1) + 1 into loongarch_rtx_cost_optimize_size instead of
> > special casing it in loongarch_rtx_costs.
>=20
> Oh yes slightly penalty is definitely not going make a huge difference if
> the cost of an mult instruction is worse than an and and an neg.
>=20
> >=20
> > For the default value (used when -O2) I'll do some micro-benchmark...

I've changed it to

/* Default RTX cost initializer.  */
loongarch_rtx_cost_data::loongarch_rtx_cost_data ()
  : fp_add (COSTS_N_INSNS (5)),
    fp_mult_sf (COSTS_N_INSNS (5)),
    fp_mult_df (COSTS_N_INSNS (5)),
    fp_div_sf (COSTS_N_INSNS (8)),
    fp_div_df (COSTS_N_INSNS (8)),
    int_mult_si (COSTS_N_INSNS (4)),
    int_mult_di (COSTS_N_INSNS (4)),
    int_div_si (COSTS_N_INSNS (5)),
    int_div_di (COSTS_N_INSNS (5)),
    branch_cost (6),
    memory_latency (4) {}

based on micro-benchmark results.  This fixes the int * _Bool case and int =
* 17
case.  But for the original test case I still get a multiplication instruct=
ion.=