Re: [Algorithms] lock-free stack with hazard list performance

[Nicholas \Indy\ R. <ar...@gm...>]

On Tue, Apr 21, 2009 at 4:28 PM, Stephan Rose <ke...@so...> wrote:
> 2. Remove Node A and delete it. Major crash right there when thread 1
> now tries do it's CAS.

Here is where I am confused, shouldn't the CAS be on the head pointer
during a dequeue so, that if the CAS is successful (assuming that
dequeue is the only way to get a reference to an item on the list,
which is how lock-free queues should be anyways) you know that you are
the only thread that has access to the node?

> Both problems can be solved by ensuring that Node A is not recycled nor
> deleted until all threads release their reference to it. All lock-free
> implementations so far I've found have been in managed languages such as
> C# or Java for that reason so far as the GC ensures exactly that.

Again, not sure why multiple threads have reference to nodes still on the queue.

Nicholas "Indy" Ray