[Bug middle-end/117338] New: VRP changing `(a & 1) == 0` into `(bool)~a` gets in the way of ccmp generation

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117338

            Bug ID: 117338
           Summary: VRP changing `(a & 1) == 0` into `(bool)~a` gets in
                    the way of ccmp generation
           Product: gcc
           Version: 15.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: enhancement
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---

Take:
```
void doit(void);

void f(int a, int b)
{
  a &= 1;
  if (a == 0 && b > 0)
    doit();
}
```

At -O1 (or -O2 -fno-tree-vrp) GCC produces:
```
f:
        tst     x0, 1
        ccmp    w1, 0, 4, eq
        bgt     .L4
```

But then at -O2 GCC produces:
```
f:
        cmp     w1, 0
        cset    w1, gt
        bics    wzr, w1, w0
        bne     .L4
```

This is because the ccmp generation code does not realize:
```
  _9 = ~a_5(D);
  _1 = (_Bool) _9;
```

is the same as `(a_5 & 1) == 1`.

The same is true of just:
```
void f(int a, int b)
{
  a &= 1;
  if (a != 0 && b > 0)
    doit();
}
```

Which has the same issue too but does not have the ! part to it.

Note this is related to PR 94174's test1 but not the same issue.