[Bug tree-optimization/80770] suboptimal code negating a 1-bit _Bool field

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80770

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
For aarch64 we get this from combine:
```
Trying 8, 10 -> 12:
    8: r92:SI=r96:QI#0&0x1
   10: r93:SI=r92:SI^0x1
      REG_DEAD r92:SI
   12: zero_extract(r99:QI#0,0x1,0)=r93:SI
      REG_DEAD r93:SI
Failed to match this instruction:
(set (zero_extract:SI (subreg:SI (reg:QI 99 [ s_4(D)->b1D.4388 ]) 0)
        (const_int 1 [0x1])
        (const_int 0 [0]))
    (not:SI (subreg:SI (reg:QI 96 [ *s_4(D) ]) 0)))
```

But I don't understand why it does not try to see we could get this:
(set (reg:SI 93) (not:SI (subreg:SI (reg:QI 96 [ *s_4(D) ]) 0)))
(set (zero_extract:SI (subreg:SI (reg:QI 99 [ s_4(D)->b1D.4388 ]) 0)
        (const_int 1 [0x1])
        (const_int 0 [0]))
    (reg:SI 93))