Re: [boost] I think using spinlocks to simulate an atomicshared_ptr isdeadlock-prone, or inefficient.

From: Talbot, George (Gtalbot_at_[hidden])
Date: 2006-11-22 16:23:43

OK...so I've been looking at this, and I'm having a bit of trouble
reading it.  Am I right in saying that ptr_base<> is your refcounted
pointer, yes?

What I'm trying to go for is to see how close to the shared_ptr<> public
interface I can get for an atomic refcounted pointer.  I'll shortly post
a working prototype wrapper for shared_ptr<> that uses a spinlock, as a
demonstration of the interface that I'm looking for.

Looking more closely at your ptr_base<> template (sorry I didn't get to
it before...I've had a bit of time the last couple of days), I would
like to discuss further with you how close in interface to shared_ptr<>
one could come with the techniques you describe.

Is the stuff (vzoom, refcount, appcore) you are working on public, or is
this part of a proprietary thing?

Thanks for pointing me to your code, it's been helpful in trying to
understand this stuff.

--
George T. Talbot
<gtalbot_at_[hidden]>
> -----Original Message-----
> From: boost-bounces_at_[hidden]
[mailto:boost-bounces_at_[hidden]]
> On Behalf Of Chris Thomasson
> Sent: Thursday, November 16, 2006 7:18 PM
> To: boost_at_[hidden]
> Subject: Re: [boost] I think using spinlocks to simulate an
> atomicshared_ptr isdeadlock-prone, or inefficient.
> 
> > I read the stuff from Chris Thommason and I have to admit that I'm
not
> > too sure how I'd wrap shared_ptr with it to make this work.
> 
> Here is an alternative to shared_ptr:
> 
> http://appcore.home.comcast.net/vzoom/refcount/
> 
> This is my version of an atomically thread-safe reference counting
that
> covers both basic and strong thread-safety models. Parts of it can
also be
> used in a signal handler, because it has 100% lock-free weak
> increments/decrements' and pointer swap operations. It amortizes lock
> actions from the first strong reference to an object X, down to the
> drop-to-zero condition of object X. What do you think?
> 
> 
> 
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost