[Bug rtl-optimization/126270] [16/17 Regression] wrong code at -O{2,3} on x86_64-linux-gnu (nondeterministic hang)

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126270

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
      Known to work|                            |15.3.0
   Target Milestone|---                         |16.2
            Summary|wrong code at -O{2,3} on    |[16/17 Regression] wrong
                   |x86_64-linux-gnu            |code at -O{2,3} on
                   |(nondeterministic hang)     |x86_64-linux-gnu
                   |                            |(nondeterministic hang)
   Last reconfirmed|                            |2026-07-15

--- Comment #2 from Drea Pinski <pinskia at gcc dot gnu.org> ---
The gimple level looks fine.
But the generated code looks to be missing something.

In GCC 15.3.0 we had:
(no zeroing the upper parts of r9b before the setne)
```
        setne   r9b
...

        je      .L8
        mov     eax, r9d
.L8:
        not     eax
        movsx   eax, al
```


While in GCC 16/17 (code gen is the same there):
```
        setne   r8b
...

        je      .L8
        mov     eax, r8d
.L8:
        not     eax
        add     eax, esi
```