4dc
Subj : Re: A question about atomic_ptr
To   : comp.programming.threads
From : Chris Thomasson
Date : Sun Apr 10 2005 06:30 am

> Can I get the best of both worlds by providing the following two methods 
> in
> shared_ptr:
>
> template<class T> class shared_ptr
> {
> // ...
>
>    shared_ptr<T> copy() const;
>    void replace( shared_ptr<T> const & pt );
> };
>
> with the guarantee that competing accesses to 'copy' and 'replace' work 
> (but
> other competing accesses are still disallowed in order to avoid an acquire
> except in 'copy')?

You can. How are you going to allow for "competing access"? Are you using 
mutexs, or going for a lock-free method such as SMR?


I am currently experimenting with using SMR to protect the reference counts 
of a quick lock-free atomic pointer API I did in C. Surprisingly, it seems 
to be more light-weight than atomic_ptr "on x86" despite the fact that it 
relies on the expensive SMR API in order to provide "strong" thread-safety. 
DWCAS can be expensive on x86. I have not compared it against Joes PPC 
reference counting algorithm that uses LL/SC directly, its very efficient.


-- 
http://appcore.home.comcast.net/
(portable lock-free data-structures)

.

0