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 119863858D1E; Sat,  3 Jun 2023 21:50:27 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 119863858D1E
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1685829027;
	bh=dkJpBACIaMNBmTltEJ7gLf0HP0SsTrgcIyoS4lnnEBk=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=s2kWZ/pfsKu6U6Ilpqx0+sakc3zShOQXBrQsirM46lNBmM+nVCrPooFtOVU7J6St3
	 JL0AIL1ovqD1uqy/1OztE1SIEeG5bJ7AGVZFwDukzxTjnQxHxXQ8WFhxj4fMS34TK/
	 +uD8ToOOlqIAi7FRh5Xtl0D391O8V686FsUfBiy4=
From: "pinskia at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug rtl-optimization/58517] ifcvt (after combine) puts ccreg
 clobbering insn between ccset insn and ccreg use
Date: Sat, 03 Jun 2023 21:50:25 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: rtl-optimization
X-Bugzilla-Version: 4.9.0
X-Bugzilla-Keywords: wrong-code
X-Bugzilla-Severity: normal
X-Bugzilla-Who: pinskia 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-58517-4-AummDtpU6J@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58517-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58517-4@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=3D58517

--- Comment #15 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I was thinking of improving this is modify noce_get_condition here (we still
need the previous patch just in case we still get the T register).

Right now noce_get_condition will stop at:
(eq (reg:SI 147 t) (const_int 0 [0]))
Because of:
  /* If the condition variable is a register and is MODE_INT, accept it.  */
  cond =3D XEXP (SET_SRC (set), 0);
  tmp =3D XEXP (cond, 0);
  if (REG_P (tmp) && GET_MODE_CLASS (GET_MODE (tmp)) =3D=3D MODE_INT
      && (GET_MODE (tmp) !=3D BImode
          || !targetm.small_register_classes_for_mode_p (BImode)))

But since 147 is a hard register but maybe since it is hard register and
REGNO_REG_CLASS (147) =3D=3D T_REGS and the register class of T_REGS is onl=
y one
register so consider a small register class.

This I think is related to the option 2 in comment #3.=