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 C77F33858D33; Tue, 20 Feb 2024 12:15:14 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org C77F33858D33
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1708431314;
	bh=iTk6Hw4eIQ1e4tr7LAVDjBBvjUdR4tM3TTrLWErdmbg=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=yk3pdMQmWmKW0tRG5j6oL5mvMnK9wDa+5W/qeqr5qrnb+P1lFUUxlf9ZhSvB2Vmzh
	 pycAt4U86pfd4rHfCs1MTrFTz4iXm4e///sGdH4ie1IlM7Sc/Eeg9aLaUIDsZDRZ3n
	 Q9RWuIV5rGz9HDvlsyGSNb0rVa35xy5cqtQ3hC3k=
From: "redi at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/109945] Escape analysis hates copy elision:
 different result with -O1 vs -O2
Date: Tue, 20 Feb 2024 12:15:13 +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: 14.0
X-Bugzilla-Keywords: wrong-code
X-Bugzilla-Severity: normal
X-Bugzilla-Who: redi 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-109945-4-fPobzSzHmb@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-109945-4@http.gcc.gnu.org/bugzilla/>
References: <bug-109945-4@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=3D109945

--- Comment #26 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #25)
> However, [class.temporary] p3 explicitly allows the compiler to create a
> temporary object when returning a class with a trivial copy constructor a=
nd
> trivial (or deleted) destructor. This is permitted precisely to allow what
> the x86_64 ABI requires: the return value is passed in registers. Complet=
ely
> eliding the creation and copy of a temporary would have required an ABI
> break.

N.B. this permission to create an intermediate temporary only depends on the
class type and whether it has a trivial copy constructor etc. *not* on the
actual platform ABI and whether it _actually_ gets passed in registers or n=
ot.

So GCC is always allowed to make a temporary copy even on i686 where the ob=
ject
is really passed in memory via invisible reference, because the class has a
trivial copy ctor.

The fact that it doesn't actually make that copy on i686 doesn't make the
program valid on i686 IMHO. Conceptually, there's a temporary which is copi=
ed
then goes out of scope, and the 'global' pointer becomes invalid and cannot=
 be
dereferenced. The fact that the actual copy isn't present any more in the
codegen (and the pointer happens to point to w itself which now exists at t=
he
same address as the temporary previously existed) doesn't make the program
valid.=