[Bug tree-optimization/114967] New: Missed optimization: std::min((int) f, -a) ==> -a where (bool f, unsigned char a)

[Bug tree-optimization/114967] New: Missed optimization: std::min((int) f, -a) ==> -a where (bool f, unsigned char a)

[652023330028 at smail dot nju.edu.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114967

            Bug ID: 114967
           Summary: Missed optimization: std::min((int) f, -a) ==> -a
                    where (bool f, unsigned char a)
           Product: gcc
           Version: 15.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: 652023330028 at smail dot nju.edu.cn
  Target Milestone: ---

Hello, we noticed that the code below can be optimized as stated in the title
(std::min((int) f, -a) ==> -a), but gcc -O3 seems to have missed it.

Reduced code:
https://godbolt.org/z/soWxqTMz8

#include <algorithm>
int n;
void func(bool f, unsigned char a) {
    n = std::min((int) f, -a);
}

GCC -O3:
  <bb 2> [local count: 1073741824]:
  # DEBUG BEGIN_STMT
  _1 = (int) a_5(D);
  _2 = -_1;
  _3 = (int) f_7(D);
  # DEBUG __a => &D.18872
  # DEBUG __b => &D.18873
  # DEBUG INLINE_ENTRY min
  if (_2 < _3)
    goto <bb 3>; [34.00%]
  else
    goto <bb 4>; [66.00%]

  <bb 3> [local count: 365072224]:

  <bb 4> [local count: 1073741824]:
  # _4 = PHI <0(2), _2(3)>
  # DEBUG __a => NULL
  # DEBUG __b => NULL
  n = _4;
  return;

func(bool, unsigned char):
        movzx   esi, sil
        movzx   edi, dil
        xor     eax, eax
        neg     esi
        cmp     esi, edi
        cmovge  esi, eax
        mov     DWORD PTR n[rip], esi
        ret


Expected code (Clang):
func(bool, unsigned char):                              # @func(bool, unsigned
char)
        neg     esi
        mov     dword ptr [rip + n], esi
        ret


Thank you very much for your time and effort! We look forward to hearing from
you.

[Bug tree-optimization/114967] Missed optimization: std::min((int) f, -a) ==> -a where (bool f, unsigned char a)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114967

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
   Last reconfirmed|                            |2024-05-07
     Ever confirmed|0                           |1
           Assignee|unassigned at gcc dot gnu.org      |pinskia at gcc dot gnu.org
             Status|UNCONFIRMED                 |ASSIGNED

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I think this is a dup of another issue I filed. Where vrp decides to replace
the value in the phi of the max and then it goes down hill from there.

I will take a look.

[Bug tree-optimization/114967] Missed optimization: std::min((int) f, -a) ==> -a where (bool f, unsigned char a)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114967

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=107888

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
We are able to handle this:
```
void func1(bool f, unsigned char a) {
    int t = f;
    int tt = -a;
    int ttt = t <= tt ? t : tt;
    n = ttt;
}
```

And we able to handle func if we disable evrp.

So yes the issue is the same as what I mentioned. Basically this is what I
mentioned in bug 107888 comment #2 . But in the original testcase in bug 107888
was able to be handled differently, it is the other testcase that still needs
to be handled.

[Bug tree-optimization/114967] Missed optimization: std::min((int) f, -a) ==> -a where (bool f, unsigned char a)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114967

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I had a patch for this which I didn't finish up and I hope to get it done for
GCC 15.