[Bug c/67090] New: [5/6 Regression] O2/O3 optimizes values away when different type pointers are typecasted

[Bug c/67090] New: [5/6 Regression] O2/O3 optimizes values away when different type pointers are typecasted

[davispuh at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67090

            Bug ID: 67090
           Summary: [5/6 Regression] O2/O3 optimizes values away when
                    different type pointers are typecasted
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: davispuh at gmail dot com
  Target Milestone: ---

GCC 5.2.0 and also latest trunk 6.0 (compiled from git commit 9b6b471, svn
r226241) using O2 or O3 generates invalid code by optimizing value away when
it's pointer is casted to different type pointer.


Here's a testcase, basically cast (double *) to (long *) and (long *) to
(double *)

#include <stdio.h>
int main(int argc, char **argv)
{
  long l = 0xAA00133700BB00EE;
  double d = (double)-2.190302448144169e-106;
  volatile int i = 0; // this to prevent optimization, could be also argc-1
  printf("%16lX\n%.15e\n", ((long*)&d)[i], ((double*)&l)[i]);
  return 0;
}

Compiling it with gcc -O1 or with clang -O3 produces correct output:

AA00133700BB00EE
-2.190302448144169e-106

But when compiling with gcc -O2 (or -O3) produces incorrect output

               0
6.952967796700164e-310


When compiled correctly with -O1

           0x004004fa    48b8ee00bb00.  movabs rax, -0x55ffecc8ff44ff12
           0x00400504    4889442418     mov qword [rsp + 0x18], rax
           0x00400509    f20f100dd700.  movsd xmm1, qword [rip + 0xd7] ;
[0x4005e8:8]=0xaa00133700bb00ee 
           0x00400511    f20f114c2410   movsd qword [rsp + 0x10], xmm1
           0x00400517    c744240c0000.  mov dword [rsp + 0xc], 0
           0x0040051f    8b54240c       mov edx, dword [rsp + 0xc]     ;
[0xc:4]=-1 ; 12
           0x00400523    8b44240c       mov eax, dword [rsp + 0xc]     ;
[0xc:4]=-1 ; 12
           0x00400527    4863d2         movsxd rdx, edx
           0x0040052a    f20f1044d418   movsd xmm0, qword [rsp + rdx*8 + 0x18]
; [0x18:8]=-1 ; 24
           0x00400530    4898           cdqe
           0x00400532    488b74c410     mov rsi, qword [rsp + rax*8 + 0x10] ;
[0x10:8]=-1 ; 16
           0x00400537    bfd4054000     mov edi, str._16lX_n_.15e_n    ;
"%16lX.%.15e." @ 0x4005d4
           0x0040053c    b801000000     mov eax, 1
           0x00400541    e89afeffff     call sym.imp.printf           
;0x004003e0() ; sym.imp.printf
           0x00400546    b800000000     mov eax, 0
           0x0040054b    4883c428       add rsp, 0x28
           0x0040054f    c3             ret

memory at rsp (0x7ffea6ee18c0)
- offset -       0 1  2 3  4 5  6 7  8 9  A B  C D  E F  0123456789ABCDEF
0x7ffea6ee18c0  5005 4000 0000 0000 1004 4000 0000 0000  P.@.......@.....
0x7ffea6ee18d0  ee00 bb00 3713 00aa ee00 bb00 3713 00aa  ....7.......7... //
values here
0x7ffea6ee18e0  0000 0000 0000 0000 9037 5a02 437f 0000  .........7Z.C...
0x7ffea6ee18f0  c819 eea6 fe7f 0000 c819 eea6 fe7f 0000  ................


When compiled incorrectly with -O2

           0x00400414    bfc4054000     mov edi, str._16lX_n_.15e_n    ;
"%16lX.%.15e." @ 0x4005c4
           0x00400419    c744240c0000.  mov dword [rsp + 0xc], 0
           0x00400421    486354240c     movsxd rdx, dword [rsp + 0xc]  ;
[0xc:4]=-1 ; 12
           0x00400426    486344240c     movsxd rax, dword [rsp + 0xc]  ;
[0xc:4]=-1 ; 12
           0x0040042b    f20f1044d410   movsd xmm0, qword [rsp + rdx*8 + 0x10]
; [0x10:8]=-1 ; 16
           0x00400431    488b74c418     mov rsi, qword [rsp + rax*8 + 0x18] ;
[0x18:8]=-1 ; 24
           0x00400436    b801000000     mov eax, 1
           0x0040043b    e8a0ffffff     call sym.imp.printf           
;0x004003e0() ; sym.imp.printf
           0x00400440    31c0           xor eax, eax
           0x00400442    4883c428       add rsp, 0x28
           0x00400446    c3             ret

memory at rsp (0x7ffd8ac419d0)
- offset -       0 1  2 3  4 5  6 7  8 9  A B  C D  E F  0123456789ABCDEF
0x7ffd8ac419d0  4005 4000 0000 0000 5004 4000 0000 0000  @.@.....P.@.....
0x7ffd8ac419e0  d01a c48a fd7f 0000 0000 0000 0000 0000  ................ //
values aren't here
0x7ffd8ac419f0  0000 0000 0000 0000 90b7 be39 047f 0000  ...........9....
0x7ffd8ac41a00  d81a c48a fd7f 0000 d81a c48a fd7f 0000  ................



Running x86_64 Arch Linux, gcc -v

COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-unknown-linux-gnu/6.0.140195.9b6b471/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: /mnt/AUR/gcc-multilib-trunk-git/src/gcc/configure
--prefix=/usr --libdir=/usr/lib --libexecdir=/usr/lib --mandir=/usr/share/man
--infodir=/usr/share/info --with-bugurl=https://bugs.archlinux.org/
--enable-languages=c,c++,lto,objc,obj-c++ --enable-shared
--enable-threads=posix --with-system-zlib --enable-__cxa_atexit
--disable-libunwind-exceptions --enable-clocale=gnu --disable-libstdcxx-pch
--disable-libssp --enable-gnu-unique-object --enable-linker-build-id
--enable-cloog-backend=isl --enable-lto --enable-plugin
--enable-install-libiberty --with-linker-hash-style=gnu --enable-multilib
--disable-werror --enable-checking=release : (reconfigured)
/mnt/AUR/gcc-multilib-trunk-git/src/gcc/configure --prefix=/usr
--libdir=/usr/lib --libexecdir=/usr/lib --mandir=/usr/share/man
--infodir=/usr/share/info --with-bugurl=https://bugs.archlinux.org/
--enable-languages=c,c++,lto,objc,obj-c++ --enable-shared
--enable-threads=posix --with-system-zlib --enable-__cxa_atexit
--disable-libunwind-exceptions --enable-clocale=gnu --disable-libstdcxx-pch
--disable-libssp --enable-gnu-unique-object --enable-linker-build-id
--enable-cloog-backend=isl --enable-lto --enable-plugin
--enable-install-libiberty --with-linker-hash-style=gnu --enable-multilib
--disable-werror --enable-checking=release
Thread model: posix
gcc version 6.0.140195.9b6b471 20150727 (experimental) (GCC)

[Bug c/67090] [5/6 Regression] O2/O3 optimizes values away when different type pointers are typecasted

[davispuh at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67090

--- Comment #1 from Dāvis <davispuh at gmail dot com> ---
Created attachment 36105
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=36105&action=edit
Testcase
>From gcc-bugs-return-493874-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Sat Aug 01 15:49:12 2015
Return-Path: <gcc-bugs-return-493874-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 108864 invoked by alias); 1 Aug 2015 15:49:12 -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 108837 invoked by uid 48); 1 Aug 2015 15:49:09 -0000
From: "pinskia at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug c/67090] [5/6 Regression] O2/O3 optimizes values away when different type pointers are typecasted
Date: Sat, 01 Aug 2015 15:49:00 -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: 6.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: pinskia at gcc dot gnu.org
X-Bugzilla-Status: RESOLVED
X-Bugzilla-Resolution: INVALID
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: bug_status resolution
Message-ID: <bug-67090-4-MDRRTamrkJ@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-67090-4@http.gcc.gnu.org/bugzilla/>
References: <bug-67090-4@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: 2015-08/txt/msg00016.txt.bz2
Content-length: 583

https://gcc.gnu.org/bugzilla/show_bug.cgi?idg090

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This is such an obvious violation of aliasing rules in c/c++.
Use memcpy or an Union (gcc extension) to get around the it. You can use
-fno-strict-aliasing too.

[Bug c/67090] [5/6 Regression] O2/O3 optimizes values away when different type pointers are typecasted

[davispuh at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67090

--- Comment #3 from Dāvis <davispuh at gmail dot com> ---
(In reply to Andrew Pinski from comment #2)
> This is such an obvious violation of aliasing rules in c/c++. 
> Use memcpy or an Union (gcc extension) to get around the it. You can use
> -fno-strict-aliasing too.

Then warning (or error) would be more appropriate than silently generating
invalid code as currently there's no warning even with -Wall
>From gcc-bugs-return-493878-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Sat Aug 01 18:33:38 2015
Return-Path: <gcc-bugs-return-493878-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 106916 invoked by alias); 1 Aug 2015 18:33:37 -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
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 106888 invoked by uid 48); 1 Aug 2015 18:33:34 -0000
From: "trippels at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug c/67090] [5/6 Regression] O2/O3 optimizes values away when different type pointers are typecasted
Date: Sat, 01 Aug 2015 18:33:00 -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: 6.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: trippels at gcc dot gnu.org
X-Bugzilla-Status: RESOLVED
X-Bugzilla-Resolution: INVALID
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: cc
Message-ID: <bug-67090-4-2igYKRkq6T@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-67090-4@http.gcc.gnu.org/bugzilla/>
References: <bug-67090-4@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: 2015-08/txt/msg00020.txt.bz2
Content-length: 883

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67090

Markus Trippelsdorf <trippels at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |trippels at gcc dot gnu.org

--- Comment #4 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
(In reply to Dāvis from comment #3)
> (In reply to Andrew Pinski from comment #2)
> > This is such an obvious violation of aliasing rules in c/c++. 
> > Use memcpy or an Union (gcc extension) to get around the it. You can use
> > -fno-strict-aliasing too.
> 
> Then warning (or error) would be more appropriate than silently generating
> invalid code as currently there's no warning even with -Wall

You'll get the warning with -Wstrict-aliasing=2 or -Wstrict-aliasing=1.
>From gcc-bugs-return-493879-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Sat Aug 01 18:38:02 2015
Return-Path: <gcc-bugs-return-493879-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 110266 invoked by alias); 1 Aug 2015 18:38:01 -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 110077 invoked by uid 55); 1 Aug 2015 18:37:57 -0000
From: "pault at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug fortran/67091] Bad result for type-bound procedures returning pointers to the intrinsic function ASSOCIATED
Date: Sat, 01 Aug 2015 18:38:00 -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: 6.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: pault at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution:
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: pault at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-67091-4-7m2ctyvBPF@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-67091-4@http.gcc.gnu.org/bugzilla/>
References: <bug-67091-4@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: 2015-08/txt/msg00021.txt.bz2
Content-length: 751

https://gcc.gnu.org/bugzilla/show_bug.cgi?idg091

--- Comment #1 from Paul Thomas <pault at gcc dot gnu.org> ---
Author: pault
Date: Sat Aug  1 18:37:25 2015
New Revision: 226464

URL: https://gcc.gnu.org/viewcvs?rev"6464&root=gcc&view=rev
Log:
2015-08-01  Paul Thomas  <pault@gcc.gnu.org>

        PR fortran/67091
        * trans-intrinsic.c (gfc_conv_associated): Add the pre and post
        blocks for the second argument to se.

2015-08-01  Paul Thomas  <pault@gcc.gnu.org>

        PR fortran/67091
        * gfortran.dg/associated_target_6.f03: New test

Added:
    trunk/gcc/testsuite/gfortran.dg/associated_target_6.f03
Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/trans-intrinsic.c
    trunk/gcc/testsuite/ChangeLog