Subj : Re: Lock Free -- where to start
To   : comp.programming.threads
From : Sean Kelly
Date : Fri Sep 30 2005 11:29 am

David Schwartz wrote:
> "Joe Seigh" <jseigh_01@xemaps.com> wrote in message
> news:AridnXxvsKImsKHeRVn-qQ@comcast.com...
>
>     No, that's not true. Blocking on a lock is going to help performance
> because useful work will be done until it's possible for the blocked thread
> to run without contention. Blocking on a lock is good because it allows
> contention to be avoided.
>
>     Lock free algorithms only help if the thing contended for is so critical
> and important that you can't do much of anything else (on the entire system)
> without accessing it. That's why they're very useful inside kernels where
> this is often true and almost useless in application where it almost never
> is.

What about deadlock resistance?  In the papers I've read, scalability
is but one of the aspects of lock-free programming that is said to be
beneficial.  I don't suppose the faq for this group covers any of these
issues?  It might help to have a clear explanation of the various uses
of locks, message-passing, and lock-free, and why one might be
preferred over another in different instances.  Probably more a topic
for a book than a few paragraphs in a faq, but it would be useful
either way.


Sean

.