[Bug target/127502] New: `neg (a ? x : 0)` -> (a ? -x : 0) not happening on the rtl level

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127502

            Bug ID: 127502
           Summary: `neg (a ? x : 0)` -> (a ? -x : 0) not happening on the
                    rtl level
           Product: gcc
           Version: 17.0
            Status: UNCONFIRMED
          Keywords: easyhack, missed-optimization
          Severity: enhancement
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---
            Target: aarch64

Take:
```
signed
f (signed x, signed tt)
{
    int t = 0;
    if (tt != 10)
      t = 1;
    return -(x * t);
}
```

Currently this forms:
```
        cmp     w1, 10
        csel    w0, w0, wzr, ne
        neg     w0, w0
```

This is because we don't prop the neg into the conditional which we can do.

That is:
```
Trying 9 -> 10:
    9: r107:SI={(cc:CC!=0)?r109:SI:0}
      REG_DEAD r109:SI
      REG_DEAD cc:CC
   10: r106:SI=-r107:SI
      REG_DEAD r107:SI
Failed to match this instruction:
(set (reg:SI 106 [ _5 ])
    (neg:SI (if_then_else:SI (ne (reg:CC 66 cc)
                (const_int 0 [0]))
            (reg:SI 109 [ xD.4512 ])
            (const_int 0 [0]))))
```

The neg can go around the `reg:SI 109`.  Though the question becomes which
version is Canonical.  So maybe this is a target issue.  Either way it is an
easy fix (add a pattern or add a simplify-rtx code to do the prop only if the
second operand is 0).