[Bug rtl-optimization/110087] New: Missing if conversion

[Bug rtl-optimization/110087] New: Missing if conversion

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

            Bug ID: 110087
           Summary: Missing if conversion
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ubizjak at gmail dot com
  Target Milestone: ---

Following testcase:

--cut here--
#include <stdbool.h>

_Bool foo (void);

_Bool bar (_Bool r)
{
  if (foo ())
    r = true;

  return r;
}
--cut here--

compiles for x86_64 target (-O2) to:

        movl    %edi, %ebx
        call    foo
        testb   %al, %al
        cmove   %ebx, %eax

More optimal code would be:

        movl    %edi, %ebx
        call    foo
        orb     %bl, %al

[Bug rtl-optimization/110087] Missing if conversion

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

--- Comment #1 from Uroš Bizjak <ubizjak at gmail dot com> ---
BTW: If the result of foo is random, then cmove gets badly predicted.
Considering the problems with cmove on x86 (even without bad prediction), the
above optimization can be quite important. Clang does it.

[Bug tree-optimization/110087] Missing if conversion

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|rtl-optimization            |tree-optimization
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |missed-optimization
   Last reconfirmed|                            |2023-06-02
                 CC|                            |pinskia at gcc dot gnu.org

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
  _Bool _1;

  <bb 2> :
  _1 = foo ();
  if (_1 != 0)
    goto <bb 3>; [INV]
  else
    goto <bb 4>; [INV]

  <bb 3> :

  <bb 4> :
  # r_2 = PHI <r_5(D)(2), 1(3)>
  return r_2;

I wonder if we should address this in PHI-OPT and transform it to

 r_2 = r_5(D) | _1;

there?  There's zero_one_valued_p we could use for both the
tested value and the value or-ed into.

We already have

/* ((x & 0x1) == 0) ? y : z <op> y -> (-(typeof(y))(x & 0x1) & z) <op> y */

and

/* ((x & 0x1) == 0) ? z <op> y : y -> (-(typeof(y))(x & 0x1) & z) <op> y */

[Bug tree-optimization/110087] Missing if conversion

[amonakov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

Alexander Monakov <amonakov at gcc dot gnu.org> changed:

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

--- Comment #3 from Alexander Monakov <amonakov at gcc dot gnu.org> ---
(In reply to Uroš Bizjak from comment #1)
> BTW: If the result of foo is random, then cmove gets badly predicted.
> Considering the problems with cmove on x86 (even without bad prediction),
> the above optimization can be quite important. Clang does it.

There is no prediction involved in execution of CMOV. It is one ALU uop with
latency 1 on any recent x86, or two uops with latency 1 on Haswell, going back
to Intel Core: https://uops.info/html-instr/CMOVZ_R32_R32.html

If you convert a control dependency to a data dependency with CMOV you may end
up with slower code due to longer dependency chains, but this is not the case
here.

[Bug tree-optimization/110087] Missing if conversion

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |pinskia at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I have a patch that converts this into the or.

[Bug tree-optimization/110087] Missing if conversion

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

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

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

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Really this is a dup of bug 89263.

*** This bug has been marked as a duplicate of bug 89263 ***

[Bug tree-optimization/110087] Missing if conversion

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #2) 
> there?  There's zero_one_valued_p we could use for both the
> tested value and the value or-ed into.
> 
> We already have
> 
> /* ((x & 0x1) == 0) ? y : z <op> y -> (-(typeof(y))(x & 0x1) & z) <op> y */
> 
> and
> 
> /* ((x & 0x1) == 0) ? z <op> y : y -> (-(typeof(y))(x & 0x1) & z) <op> y */

That is exactly what my patch does (I put it in PR 89263), I was going to
submit this patch a couple of days ago but got side tracked with other fixes
not to add the testcases to the patch and submit it.

[Bug tree-optimization/110087] Missing if conversion

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

--- Comment #7 from Uroš Bizjak <ubizjak at gmail dot com> ---
Similar conversion, not performed by gcc:

--cut here--
#include <stdbool.h>

_Bool foo (void);

int bar (int r)
{
  if (foo ())
    r++;

  return r;
}
--cut here--

gcc -O2:

        movl    %edi, %ebx
        call    foo
        cmpb    $1, %al
        sbbl    $-1, %ebx
        movl    %ebx, %eax

could be:

        movl    %edi, %ebx
        callq   foo
        movzbl  %al, %eax
        addl    %ebx, %eax

[Bug tree-optimization/110087] Missing if conversion

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Uroš Bizjak from comment #7)
> Similar conversion, not performed by gcc:
> 
> --cut here--
> #include <stdbool.h>
> 
> _Bool foo (void);
> 
> int bar (int r)
> {
>   if (foo ())
>     r++;
> 
>   return r;
> }
> --cut here--
> 
> gcc -O2:
> 
>         movl    %edi, %ebx
>         call    foo
>         cmpb    $1, %al
>         sbbl    $-1, %ebx
>         movl    %ebx, %eax
> 
> could be:
> 
>         movl    %edi, %ebx
>         callq   foo
>         movzbl  %al, %eax
>         addl    %ebx, %eax

Please file this separately, since it is a different issue. Though I think
there are dups of that one too.

[Bug tree-optimization/110087] Missing if conversion

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110087

--- Comment #9 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Andrew Pinski from comment #8)
> Please file this separately, since it is a different issue.
PR110155.