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 45ED83858D32; Sun, 12 Feb 2023 11:00:39 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 45ED83858D32
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1676199639;
	bh=k/GY+xFMlIyMT6se7f6D87rvsZNp6iN+TD6N/VurGqs=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=IVdk4Qq8XeFYcBstHBG8kP0vcI4P+oEGzchjMLLWOf+k96Nj4VTtBfLR+20ZUbpPx
	 +YLV4FpupVobPSIVGYK9+OTZLk9QP8dZ1T5WplHC1rGP41UHBPBX9uROnNALa4OpEZ
	 VYiKrnZ8pI9mHbT0S8DU5ZlUI7UGnfKVkeataPcI=
From: "sinan.lin at linux dot alibaba.com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/108764] [RISCV] Cost model for RVB is too aggressive
Date: Sun, 12 Feb 2023 11:00:38 +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: 13.0
X-Bugzilla-Keywords: 
X-Bugzilla-Severity: normal
X-Bugzilla-Who: sinan.lin at linux dot alibaba.com
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-108764-4-hlrqkWwMq5@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-108764-4@http.gcc.gnu.org/bugzilla/>
References: <bug-108764-4@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=3D108764

--- Comment #5 from Sinan <sinan.lin at linux dot alibaba.com> ---
(In reply to Kito Cheng from comment #3)
> > I think one solution is to change the cost model of such complex instru=
ctions to the sum of the cost for each part. E.g.=20
> > cost for shNadd =3D COSTS_N_INSNS (SINGLE_SHIFT_COST) + COSTS_N_INSNS (=
1) # cost of addition
>=20
> Some RISC-V core implementation did has one cycle for shNadd operation as=
 I
> know,  but I know it's not true for every implementation.

Thanks for the info. Interestingly, the shNadd-like instructions(add reg1,
reg2, reg3, lsl #N) in AArch64/neoverse-n1 are also one cycle
operations(https://developer.arm.com/documentation/pjdoc466751330-9707/late=
st),
but the cost model for them is different from the one in riscv backend(AArc=
h64
doesn't generate add r1, r2, r3, lsl #3 for the given test case).

> Anyway, it's really uarch dependent, so I would prefer keep as it for now,
> and then extend the cost model function to easier handle different uarch
> (-mtune) when GCC 14 is open.

Agree.=