[Bug tree-optimization/101025] New: wrong code at -O3 on x86_64-linux-gnu

[Bug tree-optimization/101025] New: wrong code at -O3 on x86_64-linux-gnu

[qrzhang at gatech dot edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

            Bug ID: 101025
           Summary: wrong code at -O3 on x86_64-linux-gnu
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: qrzhang at gatech dot edu
  Target Milestone: ---

It appears to be a recent regression.

Bisection points to g:128f43cf679e51564202b41f23fae4146347f93d



$ gcc-trunk -v
Supported LTO compression algorithms: zlib
gcc version 12.0.0 20210610 (experimental) [master revision
5ad76ad7f5b:c64ab1c4fe0:cce1697e6fe0134616ebc8f7040d99e66c12241a] (GCC)


$ gcc-trunk abc.c ; ./a.out
0

$ gcc-trunk -O3 abc.c ; ./a.out
4

$ cat abc.c
int a[10];
int b, d, g;
volatile char c;
short e;
volatile int f;
int main() {
  for (; d <= 9; d++) {
    b = e = 0;
    for (; e < 10; e++)
      a[e] = 4;
    for (; b <= 9; b++)
      if (g)
        f = 0;
      else
        a[b] = c;
  }
  printf("%d\n", a[1]);
}

[Bug tree-optimization/101025] [11/12 Regression] wrong code at -O3 on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |10.3.0
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
           Priority|P3                          |P2
      Known to fail|                            |11.1.0, 12.0
   Last reconfirmed|                            |2021-06-11
   Target Milestone|---                         |11.2
     Ever confirmed|0                           |1
            Version|unknown                     |12.0
            Summary|wrong code at -O3 on        |[11/12 Regression] wrong
                   |x86_64-linux-gnu            |code at -O3 on
                   |                            |x86_64-linux-gnu
             Status|UNCONFIRMED                 |ASSIGNED

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.  -fno-tree-vectorize fixes it, but in the end it seems to be
store motion that moves the vector stores across the stores in the if (!g)
block.  Thus -fno-tree-loop-im also fixes it.

[Bug tree-optimization/101025] [11/12 Regression] wrong code at -O3 on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Shorter IL testcase:

int a[10];
int b, d, g;
volatile char c;
short e;
volatile int f;
int main()
{
  for (; d <= 9; d++) {
      b = e = 0;
      for (; e < 4; e++)
        a[e] = 4;
      for (; b <= 3; b++)
        if (g)
          f = 0;
        else
          a[b] = c;
  }
  if (a[1] != 0)
    __builtin_abort ();
  return 0;
}

[Bug tree-optimization/101025] [11/12 Regression] wrong code at -O3 on x86_64-linux-gnu

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

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

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

--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
Started in between r11-4120-g94fd05f1f76faca9 and r11-4127-g35f258f4bbba7fa0, I
can't build the commit in between.

[Bug tree-optimization/101025] [11/12 Regression] wrong code at -O3 on x86_64-linux-gnu

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

--- Comment #4 from Martin Liška <marxin at gcc dot gnu.org> ---
So started with r11-4123-g128f43cf679e51564202b41f23fae4146347f93d.

[Bug tree-optimization/101025] [11/12 Regression] wrong code at -O3 on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
The issue is with sm_seq_valid_bb which when processing store sequences from
multiple branches of the CFG performs merging in a way that drops refs
that still need to be considered for dependence queries.  Notably

              /* Incrementally merge seqs into first_edge_seq.  */
              for (unsigned int i = 0; i < min_len; ++i) 
                {
                  /* ???  We can more intelligently merge when we face
different
                     order by additional sinking operations in one sequence.
                     For now we simply mark them as to be processed by the
                     not order-preserving SM code.  */
                  if (first_edge_seq[i].first != edge_seq[i].first)
                    {
                      if (first_edge_seq[i].second == sm_ord)
                        bitmap_set_bit (refs_not_supported,
                                        first_edge_seq[i].first); 
                      if (edge_seq[i].second == sm_ord)
                        bitmap_set_bit (refs_not_supported, edge_seq[i].first);
                      first_edge_seq[i].second = sm_other;
                      first_edge_seq[i].from = NULL_TREE;

we re matching f ={v} 0 against a[3] = ...;, marking 'f' as not supported for
store-motion and to be dependence checked against all stores in the sequence.
But we simply forget about a[3] = ... which we'd need to insert somehow into
the sequence for dependence checking purposes similar to how we handle

              /* Any excess elements become sm_other since they are now
                 coonditionally executed.  */
              if (first_edge_seq.length () > edge_seq.length ())
                {     

(but even that code looks suspicious).  What we'd need to verify is that
we can push the ref down in 'edge_seq' (but not actually do that) and then
append it like the late loop (but we need to disambiguate against the
stores in first_edge_seq as well).

The code is quite complex and as it handles dependence queries incrementally
it's also fragile...

[Bug tree-optimization/101025] [11/12 Regression] wrong code at -O3 on x86_64-linux-gnu

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

--- Comment #6 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:b8b80b8aa3d9a7abbcb59b651ea5e84c2ea12d0b

commit r12-1392-gb8b80b8aa3d9a7abbcb59b651ea5e84c2ea12d0b
Author: Richard Biener <rguenther@suse.de>
Date:   Fri Jun 11 12:06:08 2021 +0200

    tree-optimization/101025 - fix store-motion dependence checking

    This plugs a hole in store-motion where it fails to perform dependence
    checking on conditionally executed but not store-motioned refs.

    2021-06-11  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/101025
            * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
            all refs that require dependence checking.

            * gcc.dg/torture/pr101025.c: New testcase.

[Bug tree-optimization/101025] [11 Regression] wrong code at -O3 on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |12.0
            Summary|[11/12 Regression] wrong    |[11 Regression] wrong code
                   |code at -O3 on              |at -O3 on x86_64-linux-gnu
                   |x86_64-linux-gnu            |
      Known to fail|12.0                        |

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed on trunk sofar.  Keep more testcases coming ;)  (I have a hunch this code
has more issues...)

[Bug tree-optimization/101025] [11 Regression] wrong code at -O3 on x86_64-linux-gnu

[qrzhang at gatech dot edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

--- Comment #8 from Qirun Zhang <qrzhang at gatech dot edu> ---
(In reply to Richard Biener from comment #7)
> Fixed on trunk sofar.  Keep more testcases coming ;)  (I have a hunch this
> code has more issues...)

Thanks, Richard. Will do.

[Bug tree-optimization/101025] [11 Regression] wrong code at -O3 on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025
Bug 101025 depends on bug 101088, which changed state.

Bug 101088 Summary: [12 Regression] ICE in sm_seq_valid_bb, at tree-ssa-loop-im.c:2383
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101088

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

[Bug tree-optimization/101025] [11 Regression] wrong code at -O3 on x86_64-linux-gnu

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-11 branch has been updated by Richard Biener
<rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:8a1defec135c51531e7a0fd5085c47fca10cb147

commit r11-8695-g8a1defec135c51531e7a0fd5085c47fca10cb147
Author: Richard Biener <rguenther@suse.de>
Date:   Fri Jun 11 12:06:08 2021 +0200

    tree-optimization/101025 - fix store-motion dependence checking

    This plugs a hole in store-motion where it fails to perform dependence
    checking on conditionally executed but not store-motioned refs.

    2021-06-11  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/101025
            * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
            all refs that require dependence checking.

            * gcc.dg/torture/pr101025.c: New testcase.

    (cherry picked from commit b8b80b8aa3d9a7abbcb59b651ea5e84c2ea12d0b)

[Bug tree-optimization/101025] [11 Regression] wrong code at -O3 on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101025

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
      Known to work|                            |11.1.1
         Resolution|---                         |FIXED

--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.