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 77031385843E; Fri, 10 Mar 2023 08:50:49 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 77031385843E
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1678438249;
	bh=Gyl3EPPFJDF6WYNjvRRYPLH8A5irJOOGc7QD3AnppiY=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=hhZwOm85tto49VWXD70A5A/UjqEc/rmMHdB6fYQSjWA7LD0XTEIZzV+mubFzklDcX
	 ShM9a+LZs9wIkyH1/mZeEmX5ddamyah71nlmD9DD8QbKY+/MPjAVJ2hroC4672OGGk
	 fdLovox/EKPvurjZ1+wb5nUqoYwWqwQ1UsDCwyXU=
From: "rsandifo at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/109072] [12/13 Regression] SLP costs for vec duplicate
 too high since g:4963079769c99c4073adfd799885410ad484cbbe
Date: Fri, 10 Mar 2023 08:50:49 +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: 12.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: rsandifo at gcc dot gnu.org
X-Bugzilla-Status: ASSIGNED
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: rsandifo at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 12.3
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-109072-4-JzAem8W3xI@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-109072-4@http.gcc.gnu.org/bugzilla/>
References: <bug-109072-4@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=3D109072

--- Comment #8 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.or=
g> ---
(In reply to Richard Biener from comment #7)
> Whatever you do with cost heuristics you'll find a testcase where that
> regresses.
Yep.  That's the one true invariant of costing :-)

I think a heuristic based on vld1 does make sense though.
Code like the example is the recommended endian-agnostic way
of initialising an ACLE vector, so we should try extra hard
to vectorise stores that are later reloaded using vld1.

A longer-term fix that avoids the need for costing would be good too.
But costing seems like the easiest way of avoiding this particular
regression (in a way that's suitable for GCC 12 and 13).=