[Bug c/108482] New: ice in expand_LOOP_DIST_ALIAS, at internal-fn.cc:2737

[Bug c/108482] New: ice in expand_LOOP_DIST_ALIAS, at internal-fn.cc:2737

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

            Bug ID: 108482
           Summary: ice in expand_LOOP_DIST_ALIAS, at internal-fn.cc:2737
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dcb314 at hotmail dot com
  Target Milestone: ---

Created attachment 54317
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=54317&action=edit
C source code

For the attached C code, compiled as follows:

$ /home/dcb36/gcc/results/bin/gcc  -c -O3 -w -ftrivial-auto-var-init=zero 
bug874.c
during RTL pass: expand
testFile.21840.c: In function ‘func_1.isra’:
testFile.21840.c:112:16: internal compiler error: in expand_LOOP_DIST_ALIAS, at
internal-fn.cc:2737
0xb17ff4 expand_LOOP_DIST_ALIAS(internal_fn, gcall*)
        ../../trunk.d1/gcc/internal-fn.cc:2737
0x86d30e expand_call_stmt(gcall*)
        ../../trunk.d1/gcc/cfgexpand.cc:2737
0x86d30e expand_gimple_stmt_1(gimple*)
        ../../trunk.d1/gcc/cfgexpand.cc:3880
0x86d30e expand_gimple_stmt(gimple*)
        ../../trunk.d1/gcc/cfgexpand.cc:4044

The bug seems to exist since sometime before g:02c031088ac0bbf7,
dated 20221220.

I have a reduction running.

[Bug c/108482] ice in expand_LOOP_DIST_ALIAS, at internal-fn.cc:2737

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #1 from David Binderman <dcb314 at hotmail dot com> ---
Reduced code seems to be:

int g_30, g_261, g_263, func_1___trans_tmp_17;
int **g_120;
int *g_530;
void func_1() {
  int *l_29 = &g_30;
  *l_29 = 1;
  g_263 = 0;
  for (; g_263 <= 1; g_263 += 1) {
    g_530 = 0;
    if (*l_29) {
      char *l_1694 = &g_261;
      *l_1694 &= **g_120;
    } else
      *l_29 ^= func_1___trans_tmp_17;
  }
}

This case came from about 20,000 executions of csmith.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-01-20
   Target Milestone|---                         |13.0
            Summary|ice in                      |[13 Regression] ice in
                   |expand_LOOP_DIST_ALIAS, at  |expand_LOOP_DIST_ALIAS with
                   |internal-fn.cc:2737         |-O3
                   |                            |-ftrivial-auto-var-init=zer
                   |                            |o
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed. A regression from GCC 12.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
What is interesting is if we look at the IR:

  _41 = .LOOP_DIST_ALIAS (3, _40);
  if (_41 != 0)
    goto <bb 3>; [90.00%]
  else
    goto <bb 4>; [10.00%]

  <bb 3> [local count: 430033601]:
  _29 = *_4;
  _30 = (char) _29;
  _31 = _7 & _30;
  goto <bb 5>; [100.00%]

  <bb 4> [local count: 47781512]:
  _47 = *_4;
  _48 = (char) _47;
  _49 = _7 & _48;

  <bb 5> [local count: 118111600]:
  # _46 = PHI <_49(4), _31(3)>

That code is the same on both edges even though not noticed being the same.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
In GCC 12, IFN_LOOP_DIST_ALIAS is removed during vectorization but on the trunk
it is still there.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #5 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to David Binderman from comment #0)
> The bug seems to exist since sometime before g:02c031088ac0bbf7, dated 20221220.

I tried out a revision from a month earlier, dated 2022-11-20,
g:59cc4da605e5cb8e, and the bug seems to exist there too.

I will try 89999f2358.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #6 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to David Binderman from comment #5)
> (In reply to David Binderman from comment #0)
> > The bug seems to exist since sometime before g:02c031088ac0bbf7, dated 20221220.
> 
> I tried out a revision from a month earlier, dated 2022-11-20,
> g:59cc4da605e5cb8e, and the bug seems to exist there too.
> 
> I will try g:89999f2358724fa4

That worked, so the current range is [g:89999f2358724fa4,g:59cc4da605e5cb8e].

I will try a git bisect.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #7 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to David Binderman from comment #6)
> (In reply to David Binderman from comment #5)
> > (In reply to David Binderman from comment #0)
> > > The bug seems to exist since sometime before g:02c031088ac0bbf7, dated 20221220.
> > 
> > I tried out a revision from a month earlier, dated 2022-11-20,
> > g:59cc4da605e5cb8e, and the bug seems to exist there too.
> > 
> > I will try g:89999f2358724fa4
> 
> That worked, so the current range is [g:89999f2358724fa4,g:59cc4da605e5cb8e].
> 
> I will try a git bisect.

Git hash g:e7c8f5005bf87e4b seemed good, so I am trying g:3e070cff11f81051.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #8 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to David Binderman from comment #7)
> (In reply to David Binderman from comment #6)
> > (In reply to David Binderman from comment #5)
> > > (In reply to David Binderman from comment #0)
> > > > The bug seems to exist since sometime before g:02c031088ac0bbf7, dated 20221220.
> > > 
> > > I tried out a revision from a month earlier, dated 2022-11-20,
> > > g:59cc4da605e5cb8e, and the bug seems to exist there too.
> > > 
> > > I will try g:89999f2358724fa4
> > 
> > That worked, so the current range is [g:89999f2358724fa4,g:59cc4da605e5cb8e].
> > 
> > I will try a git bisect.
> 
> Git hash g:e7c8f5005bf87e4b seemed good, so I am trying g:3e070cff11f81051.

That seems bad, so trying g:51a233b97cc9c51e.

Current range is about 151 revisions.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #9 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to David Binderman from comment #8)
> Current range is about 151 revisions.

After a few more rounds, current range seems to be g:fbad7a74aaaddea3
to g:c16c40808331a029, some 10 commits.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

David Binderman <dcb314 at hotmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenther at suse dot de

--- Comment #10 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to David Binderman from comment #9)
> (In reply to David Binderman from comment #8)
> > Current range is about 151 revisions.
> 
> After a few more rounds, current range seems to be g:fbad7a74aaaddea3
> to g:c16c40808331a029, some 10 commits.

Both remaining revisions are by Richard, so adding them in.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #11 from David Binderman <dcb314 at hotmail dot com> ---
It looks to me like g:af96500eea72c674a5686b35c66202ef2bd9688f
is the culprit.

Over to Richard for their best advice.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This is just a latent bug exposed by Richi's copyprop improvement.


Folding statement: ivtmp_37 = ivtmp_38 - 1;
gimple_simplified to ivtmp_37 = 1;
Folded into: ivtmp_37 = 1;

Folding statement: if (ivtmp_37 != 0)
Not folded

While trunk (GCC 13):
Folding statement: ivtmp_9 = ivtmp_39 - 1;
Queued stmt for removal.  Folds to: 1
Folding statement: if (ivtmp_9 != 0)
gimple_simplified to if (1 != 0)
Folded into: if (1 != 0)

Which is the correct thing to do and removes an branch earlier.
Which then removes a few extra loops.
Except now LOOP_DIST_ALIAS is not conditionalizing two loops any more.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

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

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

--- Comment #13 from Richard Biener <rguenth at gcc dot gnu.org> ---
I will have a look.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

--- Comment #14 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Richard Biener <rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:ad4f8c4e375dd7603382ed3ff95bd184e1959918

commit r13-5300-gad4f8c4e375dd7603382ed3ff95bd184e1959918
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Jan 23 10:22:38 2023 +0100

    tree-optimization/108482 - remove stray .LOOP_DIST_ALIAS calls

    The following deals with .LOOP_DIST_ALIAS surviving vectorization
    because any of the loops involved were elided between loop distribution
    and vectorization.  As opposed to .LOOP_VECTORIZED which exists only
    between if-conversion and vectorization with no intermediate passes
    this is more difficult to deal with in advance and thus cleaning
    up after vectorization looks better.  There's the unconditional
    vector lowering pass which looks like a good place for this (for
    SIMD uid we have pass_simduid_cleanup).

            PR tree-optimization/108482
            * tree-vect-generic.cc (expand_vector_operations): Fold remaining
            .LOOP_DIST_ALIAS calls.

            * gcc.dg/torture/pr108482.c: New testcase.

[Bug tree-optimization/108482] [13 Regression] ice in expand_LOOP_DIST_ALIAS with -O3 -ftrivial-auto-var-init=zero

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108482

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

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

--- Comment #15 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.