[Bug tree-optimization/102892] New: [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[Bug tree-optimization/102892] New: [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[theodort at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

            Bug ID: 102892
           Summary: [12 Regression] Dead Code Elimination Regression at
                    -O3 (trunk vs 11.2.0)
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: theodort at inf dot ethz.ch
  Target Milestone: ---

cat test.c
static long b[2][1] = {0};
void bar(void);
void foo(void);
int main() {
  long c = 0;
  for (long a; a < 1; ++a)
    for (; c <= 1; c++) {
      bar();
      if (1 == b[c][0])
        foo();
    }
}

11.2.0 at -O3 can eliminate the call to foo but trunk at -O3 cannot:

gcc-11 -O3 -S test.c -o /dev/stdout
...
main:
.LFB0:
        .cfi_startproc
        subq    $8, %rsp
        .cfi_def_cfa_offset 16
        call    bar
        call    bar
        xorl    %eax, %eax
        addq    $8, %rsp
        .cfi_def_cfa_offset 8
        ret
        .cfi_endproc

gcc-trunk-O3 -S test.c -o /dev/stdout
main:
.LFB0:
        .cfi_startproc
        pushq   %rbx
        .cfi_def_cfa_offset 16
        .cfi_offset 3, -16
        xorl    %ebx, %ebx
        call    bar
        cmpq    $1, b(,%rbx,8)
        je      .L11
        .p2align 4,,10
        .p2align 3
.L3:
        cmpq    $1, %rbx
        je      .L12
.L5:
        call    bar
        movl    $1, %ebx
        cmpq    $1, b(,%rbx,8)
        jne     .L3
.L11:
        call    foo
        cmpq    $1, %rbx
        jne     .L5
.L12:
        xorl    %eax, %eax
        popq    %rbx
        .cfi_def_cfa_offset 8
        ret
        .cfi_endproc

gcc-trunk -v
Using built-in specs.
Target: x86_64-pc-linux-gnu
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 12.0.0 20211022 (experimental) (GCC)

Introduced with
https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=01b5038718056b024b370b74a874fbd92c5bbab3

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

--- Comment #1 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
Maybe a duplicate of PR102879?

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
   Target Milestone|---                         |12.0

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-10-23
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1

--- Comment #2 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
(In reply to Aldy Hernandez from comment #1)
> Maybe a duplicate of PR102879?

Hmmm, maybe not.

It doesn't look like a threading thing after a quick glance.  We only thread 1
path, in both 11.2 and trunk:

a.c.193t.dom3:  [4] Registering jump thread: (11, 3) incoming edge;  (3, 5)
joiner (5, 12) normal; 

And it's actually in DOM which has been largely untouched in this release.

Could someone bisect this?

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |marxin at gcc dot gnu.org
             Status|NEW                         |WAITING

--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
Well, even GCC 4.8.0 ends with:

/usr/bin/ld.bfd: /tmp/cch530u7.o: in function `main':
pr102892.c:(.text.startup+0x5): undefined reference to `bar'
/usr/bin/ld.bfd: pr102892.c:(.text.startup+0xa): undefined reference to `bar'
collect2: error: ld returned 1 exit status

What exactly should I bisect?

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

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

--- Comment #4 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
There's no call to foo in current trunk.

It seems to be gone in the .ccp4 dump which runs after vrp2.  This probably
went away with VRP2's switch to ranger.

Ideally we should add a testcase to the suite.  Could someone take care of
this?

Anywhoo, fixed.

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|needs-bisection             |

--- Comment #5 from Martin Liška <marxin at gcc dot gnu.org> ---
Fixed on master with r12-4790-g4b3a325f07acebf4.

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

--- Comment #6 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
(In reply to Martin Liška from comment #5)
> Fixed on master with r12-4790-g4b3a325f07acebf4.

Wadayaknow...I fixed it and didn't even know it :)

Thanks.

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by H.J. Lu <hjl@gcc.gnu.org>:

https://gcc.gnu.org/g:b83705b477858ba683b99fa61ff9ff83bc5be265

commit r12-5135-gb83705b477858ba683b99fa61ff9ff83bc5be265
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Wed Nov 10 11:24:44 2021 -0800

    Add a testcase for PR tree-optimization/102892

    PR tree-optimization/102892 is fixed by

    commit 4b3a325f07acebf47e82de227ce1d5ba62f5bcae
    Author: Aldy Hernandez <aldyh@redhat.com>
    Date:   Thu Oct 28 15:35:21 2021 +0200

        Remove VRP threader passes in exchange for better threading pre-VRP.

            PR tree-optimization/102892
            * gcc.dg/pr102892-1.c: New file.
            * gcc.dg/pr102892-2.c: Likewise.

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ebotcazou at gcc dot gnu.org
         Resolution|FIXED                       |---
             Status|RESOLVED                    |REOPENED

--- Comment #8 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
The test fails (to link) on SPARC though, here's the .optimized dump at -O3:

;; Function main (main, funcdef_no=0, decl_uid=1488, cgraph_uid=1,
symbol_order=1) (executed once)

Removing basic block 8
Removing basic block 9
Removing basic block 10
Removing basic block 11
int main ()
{
  unsigned int ivtmp.8;
  long int a;
  long int _1;
  void * _2;
  unsigned int _17;

  <bb 2> [local count: 44232128]:
  if (a_9(D) != 1)
    goto <bb 6>; [89.00%]
  else
    goto <bb 7>; [11.00%]

  <bb 3> [local count: 715863673]:
  # ivtmp.8_23 = PHI <ivtmp.8_24(5), ivtmp.8_7(6)>
  bar ();
  _2 = (void *) ivtmp.8_23;
  _1 = MEM[(long int *)_2];
  if (_1 == 1)
    goto <bb 4>; [20.24%]
  else
    goto <bb 5>; [79.76%]

  <bb 4> [local count: 144890806]:
  foo ();

  <bb 5> [local count: 715863674]:
  ivtmp.8_24 = ivtmp.8_23 + 4;
  if (_17 != ivtmp.8_24)
    goto <bb 3>; [66.67%]
  else
    goto <bb 7>; [33.33%]

  <bb 6> [local count: 238597362]:
  ivtmp.8_7 = (unsigned int) &b;
  _17 = ivtmp.8_7 + 8;
  goto <bb 3>; [100.00%]

  <bb 7> [local count: 44232131]:
  return 0;
}

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

--- Comment #9 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
This is an invalid test.  The variable "a" is being used before being set.  I
think we should actually remove the test from the testsuite.

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

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

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

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Guess an important question is if the #c0 testcase has been reduced from
something larger and if yes, whether it was always encountering UB at runtime
in the unreduced test as well or not.

[Bug tree-optimization/102892] [12 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[jiawei at iscas dot ac.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

jiawei <jiawei at iscas dot ac.cn> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jiawei at iscas dot ac.cn

--- Comment #11 from jiawei <jiawei at iscas dot ac.cn> ---
(In reply to Eric Botcazou from comment #8)
I use the RISC-V target test and get the same result, and link fail with
'undefined reference to `foo`'.

> The test fails (to link) on SPARC though, here's the .optimized dump at -O3:
> 
> ;; Function main (main, funcdef_no=0, decl_uid=1488, cgraph_uid=1,
> symbol_order=1) (executed once)
> 
> Removing basic block 8
> Removing basic block 9
> Removing basic block 10
> Removing basic block 11
> int main ()
> {
>   unsigned int ivtmp.8;
>   long int a;
>   long int _1;
>   void * _2;
>   unsigned int _17;
> 
>   <bb 2> [local count: 44232128]:
>   if (a_9(D) != 1)
>     goto <bb 6>; [89.00%]
>   else
>     goto <bb 7>; [11.00%]
> 
>   <bb 3> [local count: 715863673]:
>   # ivtmp.8_23 = PHI <ivtmp.8_24(5), ivtmp.8_7(6)>
>   bar ();
>   _2 = (void *) ivtmp.8_23;
>   _1 = MEM[(long int *)_2];
>   if (_1 == 1)
>     goto <bb 4>; [20.24%]
>   else
>     goto <bb 5>; [79.76%]
> 
>   <bb 4> [local count: 144890806]:
>   foo ();
> 
>   <bb 5> [local count: 715863674]:
>   ivtmp.8_24 = ivtmp.8_23 + 4;
>   if (_17 != ivtmp.8_24)
>     goto <bb 3>; [66.67%]
>   else
>     goto <bb 7>; [33.33%]
> 
>   <bb 6> [local count: 238597362]:
>   ivtmp.8_7 = (unsigned int) &b;
>   _17 = ivtmp.8_7 + 8;
>   goto <bb 3>; [100.00%]
> 
>   <bb 7> [local count: 44232131]:
>   return 0;
> }

[Bug tree-optimization/102892] [12/13 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[palmer at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

palmer at gcc dot gnu.org changed:

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

--- Comment #12 from palmer at gcc dot gnu.org ---
I just posted a patch
<https://gcc.gnu.org/pipermail/gcc-patches/2022-May/593995.html> that removes
the undefined behavior from this test case, with that it links on RISC-V.

[Bug tree-optimization/102892] [12/13 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[palmer at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

--- Comment #13 from palmer at gcc dot gnu.org ---
I just posted a patch
<https://gcc.gnu.org/pipermail/gcc-patches/2022-May/593995.html> that removes
the undefined behavior from this test case, with that it links on RISC-V.

[Bug tree-optimization/102892] [12/13 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[jiawei at iscas dot ac.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

--- Comment #14 from jiawei <jiawei at iscas dot ac.cn> ---
(In reply to palmer from comment #13)
> I just posted a patch
> <https://gcc.gnu.org/pipermail/gcc-patches/2022-May/593995.html> that
> removes the undefined behavior from this test case, with that it links on
> RISC-V.

LGTM, it's really help, thanks a lot.

[Bug tree-optimization/102892] [12/13 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|12.0                        |12.2

--- Comment #15 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 12.1 is being released, retargeting bugs to GCC 12.2.

[Bug tree-optimization/102892] [12/13 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |testsuite-fail

--- Comment #16 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to palmer from comment #13)
> I just posted a patch
> <https://gcc.gnu.org/pipermail/gcc-patches/2022-May/593995.html> that
> removes the undefined behavior from this test case, with that it links on
> RISC-V.

Can you push the patch please?  Keeping open for the testsuite FAIL for now,
please close when fixed.  The DCE regression should be INVALID after that.

[Bug tree-optimization/102892] [12/13 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

--- Comment #18 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Andrew Macleod <amacleod@gcc.gnu.org>:

https://gcc.gnu.org/g:845ee38e9b40230fbb1b9825fb62146fd858cd96

commit r13-2970-g845ee38e9b40230fbb1b9825fb62146fd858cd96
Author: Andrew MacLeod <amacleod@redhat.com>
Date:   Tue Sep 27 18:42:33 2022 -0400

    Remove undefined behaviour from testscase.

    There was a patch posted to remove the undefined behaviour from this
    testcase, but it appear to never have been applied.

            gcc/teststuite/
            PR tree-optimization/102892
            * gcc.dg/pr102892-1.c: Remove undefined behaviour.

[Bug tree-optimization/102892] [12/13 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

--- Comment #19 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-12 branch has been updated by Richard Biener
<rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:e8bd12887e5f242bc45c9df14384db052e885d80

commit r12-8847-ge8bd12887e5f242bc45c9df14384db052e885d80
Author: Andrew MacLeod <amacleod@redhat.com>
Date:   Tue Sep 27 18:42:33 2022 -0400

    Remove undefined behaviour from testscase.

    There was a patch posted to remove the undefined behaviour from this
    testcase, but it appear to never have been applied.

            gcc/teststuite/
            PR tree-optimization/102892
            * gcc.dg/pr102892-1.c: Remove undefined behaviour.

[Bug tree-optimization/102892] [12/13 Regression] Dead Code Elimination Regression at -O3 (trunk vs 11.2.0)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102892

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

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

--- Comment #20 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.