From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugs-return-441558-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 25530 invoked by alias); 26 Jan 2014 13:24:59 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Received: (qmail 25417 invoked by uid 48); 26 Jan 2014 13:24:54 -0000
From: "glisse at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug ipa/59948] Optimize std::function
Date: Sun, 26 Jan 2014 13:24:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: ipa
X-Bugzilla-Version: 4.9.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: glisse at gcc dot gnu.org
X-Bugzilla-Status: UNCONFIRMED
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-59948-4-u8M3F8sl3H@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-59948-4@http.gcc.gnu.org/bugzilla/>
References: <bug-59948-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2014-01/txt/msg02700.txt.bz2

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59948

--- Comment #1 from Marc Glisse <glisse at gcc dot gnu.org> ---
(In reply to Marc Glisse from comment #0)
>   if (f != 0B)
> // Shouldn't we know that f!=0? It is defined just above.

This happens because the test in fold-const.c is:
          return !VAR_OR_FUNCTION_DECL_P (base) || !DECL_WEAK (base);
and f is marked as weak. If I remove 'inline' or put 'static' then this code is
folded to true. However, I can't easily do the same for _M_manager. The
condition in fold-const.c probably needs to be weakened (no pun), as I strongly
doubt it is legal to replace an inline f with 0.

Note that cp/class.c contains this:
          /* ??? Probably should check DECL_WEAK here.  */
          if (t && DECL_P (t))
            *nonnull = 1;

which we may want to keep in sync.

With f static, we get the shorter but still not so good:

int m() ()
{
  struct function h;
  int _21;

  <bb 2>:
  MEM[(int (*<T58fa>) (int) *)&h] = f;
  h._M_invoker = _M_invoke;
  h.D.26465._M_manager = _M_manager;
  if (_M_manager == 0B)
// Can't make it non-weak
    goto <bb 3>;
  else
    goto <bb 4>;

  <bb 3>:
  std::__throw_bad_function_call ();

  <bb 4>:
  _21 = f (1);
// Did we notice the function is really f too late for inlining?
  std::_Function_base::_Base_manager<int (*)(int)>::_M_manager (&MEM[(struct
_Function_base *)&h]._M_functor, &MEM[(struct _Function_base *)&h]._M_functor,
3);
// Again, not inlined
  h ={v} {CLOBBER};
  h ={v} {CLOBBER};
//Maybe we could remove one of the clobbers somewhere along the way?
  return _21;

}

If I change fold-const.c (the condition is probably wrong, I just wanted to
move forward):
-  return !VAR_OR_FUNCTION_DECL_P (base) || !DECL_WEAK (base);
+  return !VAR_OR_FUNCTION_DECL_P (base) || !DECL_WEAK (base) ||
DECL_DECLARED_INLINE_P (base);

It removes the test _M_manager == 0, but it still inlines neither f nor
_M_manager.