[Bug tree-optimization/111766] New: Missed optimization with __builtin_unreachable and ands

[Bug tree-optimization/111766] New: Missed optimization with __builtin_unreachable and ands

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111766

            Bug ID: 111766
           Summary: Missed optimization with __builtin_unreachable and
                    ands
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: enhancement
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---

Take:
```
int
foo3n(int c, int bb)
{
  if ((bb & ~3)!=0) __builtin_unreachable(); // bb = [0,3]
  if ((bb & 1)==0) __builtin_unreachable(); // bb&1 == 0 // [0],[3]
  if(bb == 2) __builtin_trap();
  return bb;
}
```

The condition `bb == 2` is never true as (bb&1) has to be zero.

I Noticed this while looking into PR 111432.  Note clang/LLVM is able to
optimize this ...

[Bug tree-optimization/111766] Missed optimization with __builtin_unreachable and ands

[amacleod at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111766

--- Comment #1 from Andrew Macleod <amacleod at redhat dot com> ---


Imports: bb_3(D)
Exports: _2  bb_3(D)
         _2 : bb_3(D)(I)
bb_3(D) [irange] int [0, 3] MASK 0x3 VALUE 0x0
    <bb 4> :
    _2 = bb_3(D) & 1;
    if (_2 == 0)
      goto <bb 5>; [INV]
    else
      goto <bb 6>; [INV]

_2 : [irange] int [0, 1] MASK 0x1 VALUE 0x0
4->5  (T) _2 :  [irange] int [0, 0] MASK 0x1 VALUE 0x0
4->5  (T) bb_3(D) :     [irange] int [0, 0][2, 2] MASK 0x2 VALUE 0x0
4->6  (F) _2 :  [irange] int [1, 1] MASK 0x1 VALUE 0x0
4->6  (F) bb_3(D) :     [irange] int [1, 3] MASK 0x2 VALUE 0x1

Looks like its just a lack of completeness in
operator_bitwise_and::op1_range().
on the edge from 4->6 ranger knows _2 is [1,1], but when op1_range solves for
  _2 = bb_3 & 1
we have  
  [1,1] =  int [0, 3] MASK 0x3 VALUE 0x0    &   1
and it produces int [1, 3] rather than [1,1][3,3]

[Bug tree-optimization/111766] Missed optimization with __builtin_unreachable and ands

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111766

--- Comment #2 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:7ab79a40b546a1470abaf76bec74c63e9990fe47

commit r14-5110-g7ab79a40b546a1470abaf76bec74c63e9990fe47
Author: Andrew MacLeod <amacleod@redhat.com>
Date:   Wed Oct 25 09:46:50 2023 -0400

    Adjust operators equal and not_equal to check bitmasks against constants

    Check to see if a comparison to a constant can be determined to always
    be not-equal based on the bitmask.

            PR tree-optimization/111766
            gcc/
            * range-op.cc (operator_equal::fold_range): Check constants
            against the bitmask.
            (operator_not_equal::fold_range): Ditto.
            * value-range.h (irange_bitmask::member_p): New.

            gcc/testsuite/
            * gcc.dg/pr111766.c: New.

[Bug tree-optimization/111766] Missed optimization with __builtin_unreachable and ands

[amacleod at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111766

--- Comment #3 from Andrew Macleod <amacleod at redhat dot com> ---
fixed

[Bug tree-optimization/111766] Missed optimization with __builtin_unreachable and ands

[amacleod at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111766

Andrew Macleod <amacleod at redhat dot com> changed:

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

--- Comment #4 from Andrew Macleod <amacleod at redhat dot com> ---
oops. fixed

[Bug tree-optimization/111766] Missed optimization with __builtin_unreachable and ands

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111766

Sam James <sjames at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0