Re: [PATCH memory-barriers.txt 2/7] documentation: Fix missed renaming: s/lock/acquire

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Wed, Apr 13, 2016 at 02:46:51PM +0200, Peter Zijlstra wrote:
> On Tue, Apr 12, 2016 at 08:52:50AM -0700, Paul E. McKenney wrote:
> > From: SeongJae Park <sj38.park@gmail.com>
> > 
> > Terms `lock` and `unlock` have changed to `acquire` / `release` by
> > commit 2e4f5382d12a441b5cccfdde00308df15c2ce300 ("locking/doc: Rename
> > LOCK/UNLOCK to ACQUIRE/RELEASE").  However, the commit missed to change
> > the table of content.  This commit changes the missed parts.
> > Also, section name `Acquiring functions` is not appropriate for the
> > section because the section is saying about lock in actual.  This commit
> > changes the name to more appropriate name, `Lock acquisition functions`.
> 
> True, because of this ppc thing :/
> 
> If we get PPC to switch to RCsc locks, there actually is a difference
> again.

On that, I must defer to Michael Ellerman.

> Given the current state I'm not sure how much we should care, but
> there's a fundamental difference between things like load-acquire and
> acquiring a lock, in that the lock-acquire must also very much imply a
> store.

Agreed, even given PPC's current lock implementation, load-acquire
and lock-acquire are at best similar, not identical.  That said, one
strong similarity is the effect on ordering.

> In any case, these are jet-lagged ramblings, feel free to ignore :-)

I think we went in opposite directions.  I was in UK last week.  ;-)

							Thanx, Paul