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 3D8FE3858C78; Wed, 22 May 2024 08:26:23 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 3D8FE3858C78
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1716366385;
	bh=gepl1s5K2lmiRdod1d/q7atW8hB6ktX5FZX/NM9u7yI=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=dIgqBN1YVFhHoThUFUWZejfgoGISmhenRALC5EnzmeBT2e+1tBqq7F0MJkqQDQWQs
	 4QrizPF22nyaZ+j3BTdBsZquzDlSS8N70nEJk9riYEdscqkx/TvvD5fdfkNoUNHdDd
	 /D2wFWbuw/0+tputfdAcXETT2BbU8UWfpUK58YoE=
From: "jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/115161] [15 Regression] highway-1.0.7 miscompilation of
 some SSE2 intrinsics
Date: Wed, 22 May 2024 08:26:22 +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: 15.0
X-Bugzilla-Keywords: wrong-code
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: 15.0
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-115161-4-Es8ibTttkv@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-115161-4@http.gcc.gnu.org/bugzilla/>
References: <bug-115161-4@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=3D115161

--- Comment #19 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Alexander Monakov from comment #18)
> No, allowing value-changing transformations under -ftrapping-math is real=
ly
> not appropriate. Invoking the intrinsic on a large floating-point value is
> not UB.

If we guarantee that we never constant fold FIX/UNSIGNED_FIX with
-ftrapping-math (we shouldn't, as the exceptions should be raised), then us=
ing
FIX/UNSIGNED_FIX is ok in
flag_trapping_math guarded patterns even if the intrinsics have some specif=
ic
value they want in those cases, because they will never be folded.
While for !flag_trapping_math, we should use UNSPEC because FIX/UNSIGNED_FIX
can be folded and would be folded to whatever the generic code decides, whi=
ch
might disagree to what the intrinsics need.=