Re: "Verifying and Optimizing Compact NUMA-Aware Locks on Weak Memory Models"

[Joel Fernandes]

On Sat, Sep 10, 2022 at 4:41 PM Hernan Luis Ponce de Leon
<hernanl.leon@xxxxxxxxxx> wrote:
>
> > You were quoting Jonas here, right?  The email doesn't make this obvious
> > because it doesn't have two levels of "> > " markings.
>
> Yes, I was quoting Jonas.
> It seems my mail client did not format the email correctly and I did not notice.
> Sorry for that.
>
> > In general, _no_ two distinct relations in the LKMM have the same propagation
> > properties.  If wmb always behaved the same way as mb, we wouldn't use two
> > separate words for them.
>
> I understand that relations with different names are intended to be different.
> What I meant was
>         "wmb gives weaker propagation guarantees than mb and because of this, liveness of qspinlock is not guaranteed in LKMM"
>

I wonder if this sort of liveness guarantee (or lack thereof) is
really a problem in practice, where writes will eventually propagate
even though they may not for a bit. Is it possible to write a liveness
test case on any hardware, or is this more in the realms of theory?
Either way, quite intriguing!

Thanks,

 - Joel