[Bug tree-optimization/105618] New: Missed loop body simplification by -O3 (trunk v.s. 10.3)

[Bug tree-optimization/105618] New: Missed loop body simplification by -O3 (trunk v.s. 10.3)

[shaohua.li at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

            Bug ID: 105618
           Summary: Missed loop body simplification by -O3 (trunk v.s.
                    10.3)
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: shaohua.li at inf dot ethz.ch
  Target Milestone: ---

For the following code, gcc truck -O3 fails to optimize away the loop body,
while gcc10.3 can. You can see from the assembly code that gcc10.3 figured out
that the variable c would be a constant after the loop while gcc-trunk didn't.

$cat a.c
static int b=4;
int c;
int main() {
  int e[5] = {1,1,1,1,1};
  for (; b >= 0; b--) {
    c = e[b];
  }
  return 0;
}
$
$gcc-trunk -O3 -S -o trunk.s a.c
$gcc-10-3  -O3 -S -o 10-3.s a.c
$wc trunk.s 10-3.s
  66  147 1005 trunk.s
  52  106  698 10-3.s
 118  253 1703 total
$
$cat 10-3.s
main:
.LFB0:
        .cfi_startproc
        endbr64
        movl    b(%rip), %eax
        testl   %eax, %eax
        js      .L2
        movl    $1, c(%rip)
        movl    $-1, b(%rip)
.L2:
        xorl    %eax, %eax
        ret
$
$cat trunk.s
main:
.LFB0:
        .cfi_startproc
        movdqa  .LC0(%rip), %xmm0
        movl    b(%rip), %eax
        movl    $1, -24(%rsp)
        movaps  %xmm0, -40(%rsp)
        testl   %eax, %eax
        js      .L2
        movslq  %eax, %rdx
        leal    -1(%rax), %ecx
        movl    -40(%rsp,%rdx,4), %edx
        je      .L3
        leal    -2(%rax), %edx
        cmpl    $1, %eax
        je      .L10
        leal    -3(%rax), %ecx
        cmpl    $2, %eax
        je      .L6
        leal    -4(%rax), %edx
        cmpl    $3, %eax
        je      .L10
.L6:
        movslq  %edx, %rdx
        movl    -40(%rsp,%rdx,4), %edx
.L3:
        movl    %edx, c(%rip)
        movl    $-1, b(%rip)
.L2:
        xorl    %eax, %eax
        ret
.L10:
        movslq  %ecx, %rcx
        movl    -40(%rsp,%rcx,4), %edx
        jmp     .L3
$

[Bug tree-optimization/105618] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
If b was not static and this was not main, there could be undefined behavior.
Gcc does not take into account the function being main  though.

[Bug tree-optimization/105618] [11/12/13 Regression] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2
            Summary|Missed loop body            |[11/12/13 Regression]
                   |simplification by -O3       |Missed loop body
                   |(trunk v.s. 10.3)           |simplification by -O3
                   |                            |(trunk v.s. 10.3)
   Target Milestone|---                         |11.4
   Last reconfirmed|                            |2022-05-17
           Keywords|                            |missed-optimization,
                   |                            |needs-bisection
      Known to fail|                            |11.3.1, 12.1.1
     Ever confirmed|0                           |1
      Known to work|                            |10.3.1
             Status|UNCONFIRMED                 |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.  GCC 10 manages to sink the load from e out of the loop while GCC 12
fails to do that.

[Bug tree-optimization/105618] [11/12/13 Regression] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
There was quite some massaging in handling of PHI uses, notably we now have

              /* For PHI nodes the block we know sth about is the incoming
block
                 with the use.  */
              if (gimple_code (use_stmt) == GIMPLE_PHI)
                {
                  /* In case the PHI node post-dominates the current insert
                     location we can disregard it.  But make sure it is not
                     dominating it as well as can happen in a CFG cycle.  */
                  if (commondom != bb
                      && !dominated_by_p (CDI_DOMINATORS, commondom, bb)
                      && dominated_by_p (CDI_POST_DOMINATORS, commondom, bb)
                      /* If the blocks are possibly within the same irreducible
                         cycle the above check breaks down.  */
                      && !((bb->flags & commondom->flags & BB_IRREDUCIBLE_LOOP)
                           && bb->loop_father == commondom->loop_father)
                      && !((commondom->flags & BB_IRREDUCIBLE_LOOP)
                           && flow_loop_nested_p (commondom->loop_father,
                                                  bb->loop_father))
                      && !((bb->flags & BB_IRREDUCIBLE_LOOP)
                           && flow_loop_nested_p (bb->loop_father,
                                                  commondom->loop_father)))
                    continue;

and already the first !dominated_by_p (CDI_DOMINATORS, commondom, bb) check
makes us not consider the sink opportunity.  That was introduced as part
of the fix for PR97330.

[Bug tree-optimization/105618] [11/12/13 Regression] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

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

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

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
I'm testing a patch.

[Bug tree-optimization/105618] [11/12/13 Regression] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

--- Comment #5 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:ebce0e9bd8d714a8607ae24331a3d842b0d11859

commit r13-550-gebce0e9bd8d714a8607ae24331a3d842b0d11859
Author: Richard Biener <rguenther@suse.de>
Date:   Tue May 17 09:45:02 2022 +0200

    tree-optimization/105618 - restore load sinking

    The PR97330 fix caused some missed sinking of loads out of loops
    the following patch re-instantiates.

    2022-05-17  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/105618
            * tree-ssa-sink.cc (statement_sink_location): For virtual
            PHI uses ignore those defining the used virtual operand.

            * gcc.dg/tree-ssa/ssa-sink-19.c: New testcase.

[Bug tree-optimization/105618] [11/12 Regression] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[11/12/13 Regression]       |[11/12 Regression] Missed
                   |Missed loop body            |loop body simplification by
                   |simplification by -O3       |-O3 (trunk v.s. 10.3)
                   |(trunk v.s. 10.3)           |
      Known to work|                            |13.0

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed on trunk sofar.

[Bug tree-optimization/105618] [11/12 Regression] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

--- Comment #7 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:374cee99d01fceb89d0929da8b38051e6c9768f0

commit r12-8403-g374cee99d01fceb89d0929da8b38051e6c9768f0
Author: Richard Biener <rguenther@suse.de>
Date:   Tue May 17 09:45:02 2022 +0200

    tree-optimization/105618 - restore load sinking

    The PR97330 fix caused some missed sinking of loads out of loops
    the following patch re-instantiates.

    2022-05-17  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/105618
            * tree-ssa-sink.cc (statement_sink_location): For virtual
            PHI uses ignore those defining the used virtual operand.

            * gcc.dg/tree-ssa/ssa-sink-19.c: New testcase.

    (cherry picked from commit ebce0e9bd8d714a8607ae24331a3d842b0d11859)

[Bug tree-optimization/105618] [11 Regression] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

--- Comment #8 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:dba0883cbbd0ce7545041be96ae75db1d577affb

commit r11-10169-gdba0883cbbd0ce7545041be96ae75db1d577affb
Author: Richard Biener <rguenther@suse.de>
Date:   Tue May 17 09:45:02 2022 +0200

    tree-optimization/105618 - restore load sinking

    The PR97330 fix caused some missed sinking of loads out of loops
    the following patch re-instantiates.

    2022-05-17  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/105618
            * tree-ssa-sink.c (statement_sink_location): For virtual
            PHI uses ignore those defining the used virtual operand.

            * gcc.dg/tree-ssa/ssa-sink-19.c: New testcase.

    (cherry picked from commit ebce0e9bd8d714a8607ae24331a3d842b0d11859)

[Bug tree-optimization/105618] [11 Regression] Missed loop body simplification by -O3 (trunk v.s. 10.3)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105618

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|11.3.1                      |11.3.0
      Known to work|                            |11.3.1
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.