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 E82A33858C53; Fri,  2 Dec 2022 15:31:07 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org E82A33858C53
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1669995067;
	bh=Anu3i0fhJL+MH7mbtj2IQg1oW5IkbV2o7yfDB73O5Aw=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=SpZ5uIc+rHhn41s+pE0FSuG7ZA5wfkPcCEiiXI4s4yL5FhsKp08Md9dw5TKr1872q
	 xAi8wshgbfBDQVwzmhYj6J+TMS90IROPh4xDvDOzZhzeR1xncUhIVTF8Dbor1sTD7M
	 SI8PuRHOSAOyYc6sCyJ+eZi5T/nUy/sMs8SnkRZk=
From: "jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/107767] [13 Regression] switch to table
 conversion happening even though using btq is better
Date: Fri, 02 Dec 2022 15:31:07 +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: 13.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: jakub 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: 13.0
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-107767-4-FI93FtnCvl@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-107767-4@http.gcc.gnu.org/bugzilla/>
References: <bug-107767-4@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=3D107767

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Martin Li=C5=A1ka from comment #7)
> However, the CFG is not collapsed and thus it fails due to:
>=20
> bool
> bit_test_cluster::is_beneficial (unsigned count, unsigned uniq)
> {
>   return (((uniq =3D=3D 1 && count >=3D 3)
> 	   || (uniq =3D=3D 2 && count >=3D 5)
> 	   || (uniq =3D=3D 3 && count >=3D 6)));
> }
>=20
> as count =3D=3D 7. and so tree-switch-conversion happens. So one can miti=
gate
> that with:
> 1) use switch statement instead of if series
> 2) reduce -param=3Dswitch-conversion-max-branch-ratio=3D that will not cr=
eate so
> big CSWTCH array
> 3) disable tree-switch-conversion pass

But that just means that either iftoswitch should happen earlier or switchc=
onv
later
so that some other pass would remove the redundancies, or iftoswitch should
perform them when it optimizes something, or switchconv should not rely on =
the
collapsing being done already and do it itself.
In the #c9 testcase, it is cleanup_tree_cfg (TODO_update_ssa) during ccp1 p=
ass
that optimizes it.  But cleanup_tree_cfg (0) during iftoswitch for firewall2
doesn't do that for some reason.=