[Bug tree-optimization/112949] New: evrp produces incorrect range for __builtin_clz

[Bug tree-optimization/112949] New: evrp produces incorrect range for __builtin_clz

[kristerw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112949

            Bug ID: 112949
           Summary: evrp produces incorrect range for __builtin_clz
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: kristerw at gcc dot gnu.org
  Target Milestone: ---

The evrp pass generates incorrect ranges for __builtin_clz when it is called
within a function named __builtin_clz. While calling it in this manner seems
questionable, two relatively recent tests in the testsuite (gcc.dg/pr100521.c
and gcc.dg/pr100790.c) suggest that gcc should handle this.

The test case gcc.dg/pr100790.c is as follows:

  __builtin_clz(int x) { x ? __builtin_clz(x) : 32; }

Compiling this for x86_64 using -O3 -fpermissive results in the evrp IR:

  Global Exported: iftmp.0_3 = [irange] int [1, 31]
  __attribute__((nothrow, leaf, const))
  int __builtin_clz (int x)
  {
    int iftmp.0_3;

    <bb 2> :
    if (x_1(D) != 0)
      goto <bb 3>; [INV]
    else
      goto <bb 4>; [INV]

    <bb 3> :
    iftmp.0_3 = __builtin_clz (x_1(D));

    <bb 4> :
    return;

  }

The range for iftmp.0_3 (which is an internal call to CFN_BUILT_IN_CLZ) should
be [0, 31], not [1, 31].

[Bug tree-optimization/112949] evrp produces incorrect range for __builtin_clz

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112949

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I think the testcase is broken rather than anything else.

[Bug tree-optimization/112949] evrp produces incorrect range for __builtin_clz

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112949

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
defining __builtin_clz makes this code undefined. Especially when it comes to
calling itself.

That is the ranges here are ok and correct but rather the usage is undefined.

gcc.dg/pr100521.c is just testing to make sure we don't ICE rather than
anything else.

Likewise for gcc.dg/pr100790.c .

[Bug tree-optimization/112949] evrp produces incorrect range for __builtin_clz

[kristerw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112949

--- Comment #3 from Krister Walfridsson <kristerw at gcc dot gnu.org> ---
The C program is obviously UB. But the optimization is done on GIMPLE, and it
is not obvious to me that the GIMPLE code is UB -- we have a function called
__builtin_clz that calls an internal function, so they are different...

[Bug tree-optimization/112949] evrp produces incorrect range for __builtin_clz

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112949

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

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

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Krister Walfridsson from comment #3)
> The C program is obviously UB. But the optimization is done on GIMPLE, and
> it is not obvious to me that the GIMPLE code is UB -- we have a function
> called __builtin_clz that calls an internal function, so they are
> different...

Not exactly. GCC does not distinguish between __builtin_* the builtin vs an
user function really.  

You mention "internal call" but __builtin_clz is not an `internal call` but
rather a call to a function decl called __builtin_clz where the decl is a
builtin call.

That is the call to __builtin_clz is the same decl as the function being
defined.

There is no undefinedness in the range here since they are exactly the same.

[Bug tree-optimization/112949] evrp produces incorrect range for __builtin_clz

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112949

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Oh the definition of __builtin_clz does not have return value so the value is
undefined to begin with.

[Bug tree-optimization/112949] evrp produces incorrect range for __builtin_clz

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112949

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=32455

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
See PR 32455 and PR 99211 on why this really should have been rejected rather
than allowing a definition of __builtin_* .

[Bug tree-optimization/112949] evrp produces incorrect range for __builtin_clz

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112949

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Specifically https://gcc.gnu.org/bugzilla/show_bug.cgi?id=32455#c4 .