Re: [PATCH v3] tools/memory-model: Document herd7 (abstract) representation

From: Paul E. McKenney
Date: Tue Jun 18 2024 - 14:59:52 EST


On Tue, Jun 18, 2024 at 08:41:15PM +0200, Andrea Parri wrote:
> > Queued, thank you!
> >
> > I added Boqun's and Hernan's Reviewed-by tags and did the usual
> > wordsmithing. Please check below to make sure that I did not mess
> > anything up.
>
> Thanks! That does look good to me.
>
> It is missing the small addition to the rmw description discussed
> earlier in the thread [1]: feel free to squash it in your commit if
> that works for you (alternatively, I can respin the entire thing
> with that, JLMK what you prefer).

Please respin and I will replace the one that I have.

I clearly should have read the chain more carefully. ;-)

> > Also, Puranjay added atomic_and()/or()/xor() and add_negative, which
> > is slated to go in to the next merge window:
> >
> > be98107ab8a5 ("tools/memory-model: Add atomic_and()/or()/xor() and add_negative")
> >
> > Would you like to add the corresponding lines to this table?
>
> atomic_and() and atomic_add_negative() (together with its variants)
> should be listed in the table.
>
> I did promise myself that I would have not done "or", "xor", "andnot"
> as well as "sub", "inc", "dec", but never say never! :-) Alternatively,
> we could perhaps add a note along the lines of
>
> The table includes "add" and "and" operations; analogous/identical
> representations for "sub", "inc", "dec", "or", "xor" and "andnot"
> operations are omitted.

I am OK either way. The second approach could be used to shrink
the "RMW ops w/ return value" section, if desired.

Thanx, Paul

> Andrea
>
> [1] https://lore.kernel.org/ZnFZPJlILp5B9scN@andrea