Subj : Re: STL Containers w/ Virtually Zero-Overhead Lock-Free Reads...
To   : (Usenet)
From : Joe Seigh
Date : Sun Sep 25 2005 11:15 am

David Abrahams wrote:
> "Chris Thomasson" <_no_spam_cristom@no_spam_comcast._net> writes:
>
>>Many existing application defined algorithms that decided to use STL
>>containers in multithreaded environments use mutexs to protect everything
>>anyway; therefore they are already compatible with the requirements of the
>>highly flexible lock-free algorithm I am using. Application defined
>>algorithm can literately pick and choose exactly which if their "read-only
>>operations" can be executed without a lock. For instance, a thread that
>>makes frequent read-only const_iteration's over a shared collection would be
>>a great candidate to "lock-free-ize"...
>
>
> I thought you were "lock-free-izing" the container.  If so, you have
> done nothing to protect the entire process of iterating over it from
> being disturbed by writes.  If not, maybe you should revise your
> description of what you're proposing.

He's proposing a readers/writers solution, i.e. write with concurrent
reads.  It's for scalability which is mostly subjective.  If you
don't think you have a scalability problem then you don't need lock-free.
If you have a single threaded application then you won't expect better
performance on a multiprocessor vs. a single processor system and you
don't need lock-free.  If you have a multi-threaded application and
you don't expect better performance the more processors you have then
you don't need lock-free.

--
Joe Seigh

When you get lemons, you make lemonade.
When you get hardware, you make software.

      [ See http://www.gotw.ca/resources/clcm.htm for info about ]
      [ comp.lang.c++.moderated.    First time posters: Do this! ]

.