From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugs-return-444654-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 16303 invoked by alias); 23 Feb 2014 18:46:37 -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 16275 invoked by uid 48); 23 Feb 2014 18:46:33 -0000
From: "glisse at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/57742] memset(malloc(n),0,n) -> calloc(n,1)
Date: Sun, 23 Feb 2014 18:46:00 -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: 4.9.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: glisse at gcc dot gnu.org
X-Bugzilla-Status: NEW
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: attachments.isobsolete attachments.created
Message-ID: <bug-57742-4-F9hlkiHfIm@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-57742-4@http.gcc.gnu.org/bugzilla/>
References: <bug-57742-4@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: 2014-02/txt/msg02411.txt.bz2

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57742

Marc Glisse <glisse at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #30981|0                           |1
        is obsolete|                            |
  Attachment #31003|0                           |1
        is obsolete|                            |

--- Comment #14 from Marc Glisse <glisse at gcc dot gnu.org> ---
Created attachment 32204
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32204&action=edit
New patch

This seems to work. It also handles the fortran example from comment #3. With a
comment before the new function and a testcase, it will be good to go to
gcc-patches.

Side note: at -O3, if I provide an inline version of operator new (see PR
59894), it handles std::vector<int>(n). However, I had to provide a simple one
(call malloc, if null throw). The one in libsupc++ is way too complicated (2
calls to malloc), and if I refactor it slightly so "malloc" only appears once,
I end up with the following. The edge probabilities are strange (malloc fails
in 95% of cases?), but mostly we have a PHI node with a single argument which
hides the fact that the variables are the same. It is far from the first time I
notice this, is there a real reason to keep those unary PHIs, or should we
optimize them more aggressively?

  p_24 = mallocD.1405 (sz_20);
  if (p_24 == 0B)
    goto <bb 7>;
  else
    goto <bb 11>;
;;    succ:       7 [95.5%]  (TRUE_VALUE,EXECUTABLE)
;;                11 [4.5%]  (FALSE_VALUE,EXECUTABLE)

;;   basic block 11, loop depth 0, count 0, freq 349, maybe hot
;;    prev block 10, next block 12, flags: (NEW, REACHABLE)
;;    pred:       10 [4.5%]  (FALSE_VALUE,EXECUTABLE)
  # PT = { D.16587 } (escaped heap)
  # ALIGN = 8, MISALIGN = 0
  # p_41 = PHI <p_24(10)>
  # .MEM_42 = VDEF <.MEM_34>
  MEM[(struct _Vector_baseD.14156 *)p_2(D)]._M_implD.15030._M_startD.15032 =
p_41;
  # PT = { D.16587 } (escaped heap)
  # ALIGN = 4, MISALIGN = 0
  _19 = p_41 + sz_20;
  # .MEM_44 = VDEF <.MEM_42>
  MEM[(struct _Vector_baseD.14156
*)p_2(D)]._M_implD.15030._M_end_of_storageD.15034 = _19;
  # .MEM_8 = VDEF <.MEM_44>
  # USE = anything 
  # CLB = anything 
  memsetD.1000 (p_41, 0, sz_20);