https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110754
--- Comment #4 from Xi Ruoyao <xry111 at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #2)
> (In reply to Xi Ruoyao from comment #1)
> > Is this a bug? The standard defines accessing volatile objects as
> > side-effects so it's not allowed to merge volatile loads, AFAIU.
>
> Yes because assume attribute is defined not to have any side effects.
>
> Confirmed.
>
> gimplifier produces:
>
> [[assume (D.2786)]]
> {
> {
> int n.0;
>
> n.0 = n;
> D.2786 = n.0 == 1;
> }
> }
>
> And then lowering produces:
> _2 = n;
> .ASSUME (_Z3bari._assume.0, _2);
>
> But really it should have passed the address of n rather than the value
> since n is volatile here .
Alright, I mistakenly believed [[assume(x)]]; is same as if (!x)
unreachable();.