Incorrect bitfield aliasing with Tree SSA

[Adam Nemet]

Adam Nemet <anemet@caviumnetworks.com> writes:
> Eric Botcazou writes:
>> This is *not* OK, you're unduly pessimizing things at the RTL level.
>
> By not marking a subset relationship between the alias set of the

By marking ...

> field and the alias set of the containing aggregate where the alias
> set of the field will never be used but only the aggregate's?  How
> exactly would this effect anything other than maybe the performance of
> the compiler?
>
> Adam