[Bug c++/58542] New: std::atomic<__int128_t>::store broken with -march=corei7

[Bug c++/58542] New: std::atomic<__int128_t>::store broken with -march=corei7

[2013.bugzilla.gcc.gnu.org at ingomueller dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

            Bug ID: 58542
           Summary: std::atomic<__int128_t>::store broken with
                    -march=corei7
           Product: gcc
           Version: 4.8.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: 2013.bugzilla.gcc.gnu.org at ingomueller dot net

The attached file shows how std::atomic<__int128_t>::store is broken when
compiled with --march=corei7. Then relevant code looks like this:

    std::atomic<__int128_t> i;
    i = -1;
    std::cout << std::hex << uint64_t(i>>64) << uint64_t(i) << std::endl;

Depending on the compile flags, the program outputs the correct or an incorrect
result:

    $ g++-4.8 -std=c++11 test.cpp -O1 -latomic -o test_right && ./test_right 
    ffffffffffffffffffffffffffffffff
    $ g++-4.8 -std=c++11 test.cpp -O1 -march=corei7 -latomic -o test_wrong &&
./test_wrong 
    0ffffffffffffffff

It looks like the upper 64bit are not stored.

I tested GCC 4.7.3 and 4.8.1, but I don't have a snapshot at hand -- sorry.

[Bug c++/58542] std::atomic<__int128_t>::store broken with -march=corei7

[2013.bugzilla.gcc.gnu.org at ingomueller dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

--- Comment #1 from Ingo Müller <2013.bugzilla.gcc.gnu.org at ingomueller dot net> ---
Created attachment 30901
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30901&action=edit
Minimal "working" example of how std::atomic<__int128_t>::store can break.
>From gcc-bugs-return-430603-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Sep 26 14:23:30 2013
Return-Path: <gcc-bugs-return-430603-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 29089 invoked by alias); 26 Sep 2013 14:23:30 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 29052 invoked by uid 48); 26 Sep 2013 14:23:27 -0000
From: "glisse at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/57223] Auto-vectorization fails for nested multiple loops depending on type of array
Date: Thu, 26 Sep 2013 14:23:00 -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: 4.7.3
X-Bugzilla-Keywords:
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-57223-4-Y3DVyz5cLl@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-57223-4@http.gcc.gnu.org/bugzilla/>
References: <bug-57223-4@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: 2013-09/txt/msg01843.txt.bz2
Content-length: 677

http://gcc.gnu.org/bugzilla/show_bug.cgi?idW223

--- Comment #5 from Marc Glisse <glisse at gcc dot gnu.org> ---
(In reply to Usishchev Yury from comment #3)
> I'm testing it on current trunk, and second loop is not vectorized both with
> floating point and integer types.
> For floating point types it is not vectorized due to control flow in loop:
>
>  <bb 15>:
> // ...
> if (t_56 > _61)
>   goto <bb 16>;
> else
>   goto <bb 17>;
>  <bb 16>:
>  <bb 17>:
> # iftmp.2_7 = PHI <_61(16), t_56(15)>
>
> This can be optimized to MIN_EXPR in phiopt pass, but is not because of NaNs:

Even without MIN_EXPR, ifcvt should turn this into a COND_EXPR that can be
vectorized, no?

[Bug rtl-optimization/58542] std::atomic<__int128_t>::store broken with -march=corei7

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-09-26
          Component|c++                         |rtl-optimization
     Ever confirmed|0                           |1

--- Comment #2 from Uroš Bizjak <ubizjak at gmail dot com> ---
Confirmed, it is subreg1 pass that splits:

(insn 10 9 11 3 (set (reg:TI 107)
        (const_double -1 [0xffffffffffffffff] 0 [0] 0 [0] 0 [0]))
/usr/include/c++/4.8.1/atomic:199 84 {*movti_internal}
     (nil))

to:

(insn 58 57 59 3 (set (reg:DI 130)
        (const_int -1 [0xffffffffffffffff])) /usr/include/c++/4.8.1/atomic:199
85 {*movdi_internal}
     (nil))
(insn 59 58 11 3 (set (reg:DI 131 [+8 ])
        (const_int 0 [0])) /usr/include/c++/4.8.1/atomic:199 85
{*movdi_internal}
     (nil))
>From gcc-bugs-return-430626-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Sep 26 17:22:54 2013
Return-Path: <gcc-bugs-return-430626-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 26859 invoked by alias); 26 Sep 2013 17:22:53 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 26823 invoked by uid 48); 26 Sep 2013 17:22:50 -0000
From: "ubizjak at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] std::atomic<__int128_t>::store broken with -march=corei7
Date: Thu, 26 Sep 2013 17:22:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: rtl-optimization
X-Bugzilla-Version: 4.8.1
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: ubizjak at gmail dot com
X-Bugzilla-Status: NEW
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 4.7.4
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields: target_milestone
Message-ID: <bug-58542-4-QWEWcthvl3@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58542-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58542-4@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
X-SW-Source: 2013-09/txt/msg01866.txt.bz2
Content-length: 294

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.4
>From gcc-bugs-return-430625-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Sep 26 17:22:41 2013
Return-Path: <gcc-bugs-return-430625-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 26182 invoked by alias); 26 Sep 2013 17:22:41 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 26146 invoked by uid 48); 26 Sep 2013 17:22:39 -0000
From: "ubizjak at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] std::atomic<__int128_t>::store broken with -march=corei7
Date: Thu, 26 Sep 2013 17:22:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: rtl-optimization
X-Bugzilla-Version: 4.8.1
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: ubizjak at gmail dot com
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: short_desc
Message-ID: <bug-58542-4-g2TkIYUi6J@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58542-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58542-4@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
X-SW-Source: 2013-09/txt/msg01865.txt.bz2
Content-length: 631

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|std::atomic<__int128_t>::st |[4.7/4.8/4.9 Regression]
                   |ore broken with             |std::atomic<__int128_t>::st
                   |-march=corei7               |ore broken with
                   |                            |-march=corei7

--- Comment #3 from Uroš Bizjak <ubizjak at gmail dot com> ---
Fails on all branches, including trunk.
>From gcc-bugs-return-430627-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Sep 26 17:44:50 2013
Return-Path: <gcc-bugs-return-430627-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 15034 invoked by alias); 26 Sep 2013 17:44:49 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 14763 invoked by uid 55); 26 Sep 2013 17:44:18 -0000
From: "edlinger at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug fortran/58113] [4.9 Regression] gfortran.dg/round_4.f90 FAILs
Date: Thu, 26 Sep 2013 17:44:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: fortran
X-Bugzilla-Version: 4.9.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: edlinger 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: 4.9.0
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-58113-4-vXTLgYb0My@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58113-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58113-4@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: 2013-09/txt/msg01867.txt.bz2
Content-length: 479

http://gcc.gnu.org/bugzilla/show_bug.cgi?idX113

--- Comment #5 from edlinger at gcc dot gnu.org ---
Author: edlinger
Date: Thu Sep 26 17:44:13 2013
New Revision: 202954

URL: http://gcc.gnu.org/viewcvs?rev 2954&root=gcc&view=rev
Log:
2013-09-26  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR fortran/58113
        * gfortran.dg/round_4.f90: Check for rounding support.


Modified:
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/gfortran.dg/round_4.f90

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] subreg splitting pass mishandles TImode immediates

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |iant at google dot com,
                   |                            |rth at gcc dot gnu.org
            Summary|[4.7/4.8/4.9 Regression]    |[4.7/4.8/4.9 Regression]
                   |std::atomic<__int128_t>::st |subreg splitting pass
                   |ore broken with             |mishandles TImode
                   |-march=corei7               |immediates

--- Comment #4 from Uroš Bizjak <ubizjak at gmail dot com> ---
Adjusted Summary, CC'd authors.
>From gcc-bugs-return-430635-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Sep 26 22:41:51 2013
Return-Path: <gcc-bugs-return-430635-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 9598 invoked by alias); 26 Sep 2013 22:41:50 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 9560 invoked by uid 48); 26 Sep 2013 22:41:47 -0000
From: "regehr at cs dot utah.edu" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug c/58546] volatile bug and also larger code at -Os
Date: Thu, 26 Sep 2013 22:41:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: c
X-Bugzilla-Version: 4.9.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: regehr at cs dot utah.edu
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-58546-4-KEAqOEhgJ1@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58546-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58546-4@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: 2013-09/txt/msg01875.txt.bz2
Content-length: 312

http://gcc.gnu.org/bugzilla/show_bug.cgi?idX546

--- Comment #2 from John Regehr <regehr at cs dot utah.edu> ---
I agree, this is not very likely to cause problems in practice, though I
selfishly hope that it gets fixed because then this bug will stop getting in
our way when we look for other volatile bugs.

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] subreg splitting pass mishandles TImode immediates

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

--- Comment #5 from Uroš Bizjak <ubizjak at gmail dot com> ---
The problem actually starts in expand_atomic_compare_and_swap, in:

(gdb) list
7339          create_convert_operand_to (&ops[3], expected, mode, true);
7340          create_convert_operand_to (&ops[4], desired, mode, true);
7341          create_integer_operand (&ops[5], is_weak);
7342          create_integer_operand (&ops[6], succ_model);
7343          create_integer_operand (&ops[7], fail_model);
7344          expand_insn (icode, 8, ops);

ops[4] is converted in unsigned mode, so from "desired" operand:

(gdb) p debug_rtx (desired)
(const_int -1 [0xffffffffffffffff])

we got:

(gdb) p ops[4]
$45 = {type = EXPAND_CONVERT_TO, unsigned_p = 1, unused = 0, mode = TImode,
value = 0x7fffeffc21e0}
(gdb) p debug_rtx (ops[4].value)
(const_double -1 [0xffffffffffffffff] 0 [0] 0 [0] 0 [0])

So, it is actually expansion of atomic_compare_and_swap, which doesn't account
for signedness of "desired" operand.

Manually changing the argument from "true" to "false" for ops[4] generates
correct code.
>From gcc-bugs-return-431315-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Tue Oct 08 16:23:05 2013
Return-Path: <gcc-bugs-return-431315-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 24222 invoked by alias); 8 Oct 2013 16:23:05 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 24135 invoked by uid 55); 8 Oct 2013 16:23:02 -0000
From: "dehao at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/58619] ICE building in gen_combined_adhoc_loc
Date: Tue, 08 Oct 2013 16:23:00 -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: unknown
X-Bugzilla-Keywords: ice-on-valid-code
X-Bugzilla-Severity: normal
X-Bugzilla-Who: dehao 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: 4.9.0
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-58619-4-jpBTOueuyc@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58619-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58619-4@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: 2013-10/txt/msg00459.txt.bz2
Content-length: 466

http://gcc.gnu.org/bugzilla/show_bug.cgi?idX619

--- Comment #5 from dehao at gcc dot gnu.org ---
Author: dehao
Date: Tue Oct  8 16:22:57 2013
New Revision: 203284

URL: http://gcc.gnu.org/viewcvs?rev 3284&root=gcc&view=rev
Log:
Backport r203269.

PR tree-optimization/58619

2013-10-08  Dehao Chen  <dehao@google.com>

    * tree-inline.c (copy_phis_for_bb): Combine location data
    only if non-null.

Modified:
    branches/google/gcc-4_8/gcc/tree-inline.c

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] subreg splitting pass mishandles TImode immediates

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |ubizjak at gmail dot com

--- Comment #6 from Uroš Bizjak <ubizjak at gmail dot com> ---
Created attachment 30968
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30968&action=edit
patch that converts arguments to __atomic builtins in signed mode

Patch in testing.
>From gcc-bugs-return-431324-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Tue Oct 08 18:25:40 2013
Return-Path: <gcc-bugs-return-431324-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 14197 invoked by alias); 8 Oct 2013 18:25:39 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 14166 invoked by uid 48); 8 Oct 2013 18:25:37 -0000
From: "ubizjak at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] Arguments of __atomic_* functions are converted in unsigned mode
Date: Tue, 08 Oct 2013 18:25:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: rtl-optimization
X-Bugzilla-Version: 4.8.1
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: ubizjak at gmail dot com
X-Bugzilla-Status: ASSIGNED
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: ubizjak at gmail dot com
X-Bugzilla-Target-Milestone: 4.7.4
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields: short_desc
Message-ID: <bug-58542-4-ae85V0nI7t@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58542-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58542-4@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
X-SW-Source: 2013-10/txt/msg00468.txt.bz2
Content-length: 618

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.7/4.8/4.9 Regression]    |[4.7/4.8/4.9 Regression]
                   |subreg splitting pass       |Arguments of __atomic_*
                   |mishandles TImode           |functions are converted in
                   |immediates                  |unsigned mode

--- Comment #7 from Uroš Bizjak <ubizjak at gmail dot com> ---
Update summary (again).
>From gcc-bugs-return-431325-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Tue Oct 08 18:38:45 2013
Return-Path: <gcc-bugs-return-431325-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 23929 invoked by alias); 8 Oct 2013 18:38:45 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 23881 invoked by uid 48); 8 Oct 2013 18:38:42 -0000
From: "ubizjak at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] Arguments of __atomic_* functions are converted in unsigned mode
Date: Tue, 08 Oct 2013 18:38:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: rtl-optimization
X-Bugzilla-Version: 4.8.1
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: ubizjak at gmail dot com
X-Bugzilla-Status: ASSIGNED
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: ubizjak at gmail dot com
X-Bugzilla-Target-Milestone: 4.7.4
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields: bug_file_loc
Message-ID: <bug-58542-4-2B9M7did52@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58542-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58542-4@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
X-SW-Source: 2013-10/txt/msg00469.txt.bz2
Content-length: 588

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |http://gcc.gnu.org/ml/gcc-p
                   |                            |atches/2013-10/msg00496.htm
                   |                            |l

--- Comment #8 from Uroš Bizjak <ubizjak at gmail dot com> ---
Patch at [1].

[1] http://gcc.gnu.org/ml/gcc-patches/2013-10/msg00496.html
>From gcc-bugs-return-431326-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Tue Oct 08 18:45:06 2013
Return-Path: <gcc-bugs-return-431326-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 30125 invoked by alias); 8 Oct 2013 18:45:06 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 30067 invoked by uid 48); 8 Oct 2013 18:45:03 -0000
From: "ebotcazou at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug middle-end/58570] [4.9 Regression] wrong code for bitfields at -O2 and above
Date: Tue, 08 Oct 2013 18:45:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: middle-end
X-Bugzilla-Version: 4.9.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: ebotcazou at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: ebotcazou at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 4.9.0
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-58570-4-gci2JuT1yj@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58570-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58570-4@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: 2013-10/txt/msg00470.txt.bz2
Content-length: 351

http://gcc.gnu.org/bugzilla/show_bug.cgi?idX570

--- Comment #12 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> No. You only assume an alias if _both_ fields are bit fields.
> But in my example only one "a" is a volatile bit field the other
> is a normal member "b".

Then they won't be affected by the bug, see my explanation to Richard.

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] Arguments of __atomic_* functions are converted in unsigned mode

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW
                URL|http://gcc.gnu.org/ml/gcc-p |
                   |atches/2013-10/msg00496.htm |
                   |l                           |
           Assignee|ubizjak at gmail dot com           |unassigned at gcc dot gnu.org

--- Comment #9 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Uroš Bizjak from comment #8)
> Patch at [1].

According to a review, the proposed fix is wrong.

I will humbly step aside and leave the honors of the proper fix to someone
else.
>From gcc-bugs-return-432547-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Wed Oct 23 12:15:09 2013
Return-Path: <gcc-bugs-return-432547-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 4772 invoked by alias); 23 Oct 2013 12:15:08 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 4719 invoked by uid 48); 23 Oct 2013 12:15:04 -0000
From: "paolo.carlini at oracle dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug c++/58822] Segfault when calling make_shared
Date: Wed, 23 Oct 2013 12:15:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: c++
X-Bugzilla-Version: 4.8.1
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: paolo.carlini at oracle dot com
X-Bugzilla-Status: REOPENED
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-58822-4-TqkNMgDHlJ@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58822-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58822-4@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: 2013-10/txt/msg01691.txt.bz2
Content-length: 630

http://gcc.gnu.org/bugzilla/show_bug.cgi?idX822

--- Comment #19 from Paolo Carlini <paolo.carlini at oracle dot com> ---
I'm pretty sure it isn't. It's easy to see why: in the testcase are you using
*all* the facilities provided by std::shared_ptr as standardized? I don't think
so! Thus just imagine having the class there, and commenting out all the member
functions you don't use. Likewise for the free functions. Also, including
<memory> means *a ton* of code which definitely has *nothing* to do with
std::shared_ptr. Normally, a C++ testcase is less than 100 lines without
includes. The exceptions are *extremely* rare.

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] Arguments of __atomic_* functions are converted in unsigned mode

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Richard Henderson <rth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |rth at gcc dot gnu.org

--- Comment #10 from Richard Henderson <rth at gcc dot gnu.org> ---
Thanks for reminding me of this.

The bug isn't in expand_atomic_compare_and_swap, and the parameters
to create_convert_operand_to, because the argument is already TImode:

 <integer_cst 0x7ffff19d75e0
   type <integer_type 0x7ffff19e1a80 __int128 unsigned>
   constant -1>

The bug is either in expand_expr_* in creating a const_int instead of
a const_double, or in convert_to_mode (or its callees) in misinterpreting
the const_int.

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] Arguments of __atomic_* functions are converted in unsigned mode

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

--- Comment #11 from Richard Henderson <rth at gcc dot gnu.org> ---
Author: rth
Date: Thu Oct 24 22:27:53 2013
New Revision: 204040

URL: http://gcc.gnu.org/viewcvs?rev=204040&root=gcc&view=rev
Log:
PR rtl/58542

* optabs.c (maybe_emit_atomic_exchange): Use create_input_operand
instead of create_convert_operand_to.
(maybe_emit_sync_lock_test_and_set): Likewise.
(expand_atomic_compare_and_swap): Likewise.
(maybe_emit_compare_and_swap_exchange_loop): Don't convert_modes.

Added:
    trunk/gcc/testsuite/gcc.dg/atomic-store-6.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/optabs.c
    trunk/gcc/testsuite/ChangeLog

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] Arguments of __atomic_* functions are converted in unsigned mode

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

--- Comment #12 from Richard Henderson <rth at gcc dot gnu.org> ---
Author: rth
Date: Fri Oct 25 20:06:20 2013
New Revision: 204076

URL: http://gcc.gnu.org/viewcvs?rev=204076&root=gcc&view=rev
Log:
PR rtl/58542

  * optabs.c (maybe_emit_atomic_exchange): Use create_input_operand
  instead of create_convert_operand_to.
  (maybe_emit_sync_lock_test_and_set): Likewise.
  (expand_atomic_compare_and_swap): Likewise.
  (maybe_emit_compare_and_swap_exchange_loop): Don't convert_modes.

Added:
    branches/gcc-4_8-branch/gcc/testsuite/gcc.dg/atomic-store-6.c
Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/optabs.c
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] Arguments of __atomic_* functions are converted in unsigned mode

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

--- Comment #13 from Richard Henderson <rth at gcc dot gnu.org> ---
Author: rth
Date: Fri Oct 25 20:09:13 2013
New Revision: 204077

URL: http://gcc.gnu.org/viewcvs?rev=204077&root=gcc&view=rev
Log:
PR rtl/58542

  * optabs.c (maybe_emit_atomic_exchange): Use create_input_operand
  instead of create_convert_operand_to.
  (maybe_emit_sync_lock_test_and_set): Likewise.
  (expand_atomic_compare_and_swap): Likewise.
  (maybe_emit_compare_and_swap_exchange_loop): Don't convert_modes.

Added:
    branches/gcc-4_7-branch/gcc/testsuite/gcc.dg/atomic-store-6.c
Modified:
    branches/gcc-4_7-branch/gcc/ChangeLog
    branches/gcc-4_7-branch/gcc/optabs.c
    branches/gcc-4_7-branch/gcc/testsuite/ChangeLog

[Bug rtl-optimization/58542] [4.7/4.8/4.9 Regression] Arguments of __atomic_* functions are converted in unsigned mode

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58542

Richard Henderson <rth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #14 from Richard Henderson <rth at gcc dot gnu.org> ---
Fixed.