[Bug rtl-optimization/97836] New: wrong code at -O1 on x86_64-pc-linux-gnu

[Bug rtl-optimization/97836] New: wrong code at -O1 on x86_64-pc-linux-gnu

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

            Bug ID: 97836
           Summary: wrong code at -O1 on x86_64-pc-linux-gnu
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zhendong.su at inf dot ethz.ch
  Target Milestone: ---

[510] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/11.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--prefix=/local/suz-local/software/local/gcc-trunk --enable-languages=c,c++
--disable-werror --enable-multilib --with-system-zlib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 11.0.0 20201115 (experimental) [master revision
406b452dc0e:d88ff5a679d:faab61b585a8b4a42454f085dd6b7815992a98f5] (GCC) 
[511] % 
[511] % gcctk -O0 small.c; ./a.out
[512] % 
[512] % gcctk -O1 small.c; ./a.out
Aborted
[513] % 
[513] % cat small.c
int *a, b;

int *f(int *c) {
  b = 0;
  return c;
}

int main() {
  int d;
  a = f(&d);
  if (a != &d)
    __builtin_abort ();
  return 0;
}

[Bug rtl-optimization/97836] wrong code at -O1 on x86_64-pc-linux-gnu

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

--- Comment #1 from Zhendong Su <zhendong.su at inf dot ethz.ch> ---
Another related test that triggers the miscompilation at -Os, but not -O1: 

[621] % gcctk -O1 small.c; ./a.out
[622] % 
[622] % gcctk -Os small.c; ./a.out
Aborted
[623] % 
[623] % cat small.c
int a;

int b(int c) { return 0; }

static int *d(int *e) {
  if (a) {
    a = a && b(*e);
  }
  return e;
}

int main() {
  int f;
  if (d(&f) != &f)
    __builtin_abort();
  return 0;
}

[Bug rtl-optimization/97836] [11 Regression] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|wrong code at -O1 on        |[11 Regression] wrong code
                   |x86_64-pc-linux-gnu         |at -O1 on
                   |                            |x86_64-pc-linux-gnu by
                   |                            |r11-5029
     Ever confirmed|0                           |1
                 CC|                            |hjl.tools at gmail dot com,
                   |                            |jh at suse dot cz
   Target Milestone|---                         |11.0
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2020-11-15

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> ---
It is caused by r11-5029

[Bug rtl-optimization/97836] [11 Regression] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
                 CC|                            |hubicka at gcc dot gnu.org

--- Comment #3 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Confirmed.
The wrong code happens already in fre1 where we do:

main ()
{
  int f;
  int * _1;

  <bb 2> :
  _1 = d (&f);
  __builtin_abort ();

}

Modref summary for d is:
  loads:
    Limits: 32 bases, 16 refs
      Base 0: alias set 1
        Ref 0: alias set 1
          Every access
  stores:
    Limits: 32 bases, 16 refs
      Base 0: alias set 1
        Ref 0: alias set 1
          Every access
  parm 0 flags: direct noclobber noescape unused

for body:

d (int * e)
{
  int D.1973;
  int a.0_1;

  <bb 2> :
  a.0_1 = a;
  if (a.0_1 != 0)
    goto <bb 3>; [INV]
  else
    goto <bb 4>; [INV]

  <bb 3> :
  a = 0;

  <bb 4> :
  return e_10(D);

}

direct noclobber noescape looks correct to me: value is only returned and
noescape values are allowed to escape to return value (per IRC discussion we
had with Richi).

I think problem is with unused that makes tree-ssa-structalias to completely
skip the parameter rather than adding it to return value alias set.

I guess we want to specify what unused really means. Indeed current comment is
"Nonzero if the argument is not used by the function." and in this case we wold
need to have separate EAF_NOREAD so current EAF_UNUSED would be EAF_NOCLOBBER |
EAF_NOREAD or track that internally in ipa-modref.

A quick fix is to make return statement clear EAF_UNUSED flag.

[Bug rtl-optimization/97836] [11 Regression] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
           Priority|P3                          |P1

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Indeed EAF_UNUSED was supposed to be really unused (which includes not escaping
anywhere, including to the return).  I guess we could change it to allow
escaping to return in case fully unused isn't useful.  Nothing used EAF_UNUSED
I think and I just added it for completeness back in time...

[Bug rtl-optimization/97836] [11 Regression] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

--- Comment #5 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
I forgot to attach the PR number, but I commited the quick fix (to prevent
wrong code) as g:26285af40f98dfdb809b98b08386073c63b65db1

I will discuss the EAF_UNUSED flag today after teaching.

[Bug rtl-optimization/97836] [11 Regression] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
So, can this be closed as fixed or do you want to do something further?

[Bug rtl-optimization/97836] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[11 Regression] wrong code  |wrong code at -O1 on
                   |at -O1 on                   |x86_64-pc-linux-gnu by
                   |x86_64-pc-linux-gnu by      |r11-5029
                   |r11-5029                    |

--- Comment #7 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
It is no longer regression,  I will look into defining EAF_UNUSED better next
stage1.

[Bug rtl-optimization/97836] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

--- Comment #8 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
indeed, I think for gcc11 we want to make return mark value as used and for
next stage1 we want to design EAF flags bit more carefully...

[Bug rtl-optimization/97836] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|11.0                        |11.2

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 11.1 has been released, retargeting bugs to GCC 11.2.

[Bug rtl-optimization/97836] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|11.2                        |11.3

--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 11.2 is being released, retargeting bugs to GCC 11.3

[Bug rtl-optimization/97836] wrong code at -O1 on x86_64-pc-linux-gnu by r11-5029

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97836

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

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

--- Comment #11 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
EAF_UNUSED is now realy unused.