[Bug c/111088] New: useless 'xor eax,eax' inserted when a value is not returned.

[Bug c/111088] New: useless 'xor eax,eax' inserted when a value is not returned.

[charly.poyac at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111088

            Bug ID: 111088
           Summary: useless 'xor eax,eax' inserted when a value is not
                    returned.
           Product: gcc
           Version: 13.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: charly.poyac at gmail dot com
  Target Milestone: ---

I compile this simple C code with GCC 13.2:

int foo(int *a)
{
    *a += 1;
}

int foo2(int *a)
{
    *a += 1;
} 

and it produces the following assembly code with '-O3':

foo:
        add     DWORD PTR [rdi], 1
        ret
foo2:
        add     DWORD PTR [rdi], 1
        xor     eax, eax
        ret

see godbolt: https://godbolt.org/z/eeT63ePcr

We can see the inconsistency between the two functions with the 'xor eax,eax'
added in foo2. This does not happen if I compile with -O1.

Yes, these functions should return an integer, but they don't. However, this is
allowed by the C standard (2018 6.9.1 11/12).

[Bug middle-end/111088] useless 'xor eax,eax' inserted when a value is not returned and icf

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111088

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
It is ICF coming into play and then inlining.

[Bug middle-end/111088] useless 'xor eax,eax' inserted when a value is not returned and icf

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111088

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2023-08-21
             Status|UNCONFIRMED                 |NEW

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
So ICF figures out foo and foo2 are the same and has foo2 tail calls foo:

  <bb 2> [local count: 1073741824]:
  retval.2_4 = foo (a_2(D)); [tail call]
  return retval.2_4;

But adds a return with a value. And then the inliner inlines foo into foo2 but
we still have the return with a value around ...

Re: [Bug middle-end/111088] useless 'xor eax,eax' inserted when a value is not returned and icf

[Jan Hubicka]

> But adds a return with a value. And then the inliner inlines foo into foo2 but
> we still have the return with a value around ...
I guess ICF can special case unused return value, but why this is not
taken care of by ipa-sra?

[Bug ipa/111088] useless 'xor eax,eax' inserted when a value is not returned and icf

[hubicka at ucw dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111088

--- Comment #3 from Jan Hubicka <hubicka at ucw dot cz> ---
> But adds a return with a value. And then the inliner inlines foo into foo2 but
> we still have the return with a value around ...
I guess ICF can special case unused return value, but why this is not
taken care of by ipa-sra?

[Bug ipa/111088] useless 'xor eax,eax' inserted when a value is not returned and icf

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111088

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
In C++ this would have been undefined - I don't think this particular testcase
is worth "optimizing".

[Bug ipa/111088] useless 'xor eax,eax' inserted when a value is not returned and icf

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111088

Martin Jambor <jamborm at gcc dot gnu.org> changed:

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

--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> ---
(In reply to Jan Hubicka from comment #3)
> > But adds a return with a value. And then the inliner inlines foo into foo2 but
> > we still have the return with a value around ...
> I guess ICF can special case unused return value, but why this is not
> taken care of by ipa-sra?

Probably because none of the functions is static.