>
> Regarding the existing standard C++ attributes, I am not quite sure that
> any of them could meaningfully be applied to expressions; maybe just
> [[likely]]?
>

I also couldn't think of much, but using [[likely]] on one branch of a
conditional operator seems pretty reasonable.

Maybe there are some GNU attributes that would provide more motivation.

I've also once seen a discussion about an inlining attribute applied to
function call expressions, which would be interesting.