Subj : Re: Optimization to Jeffrey Richter's COptex (Windows)?
To   : comp.programming.threads
From : Joseph Seigh
Date : Wed Jan 26 2005 03:03 pm

On Wed, 26 Jan 2005 20:03:37 +0100, Alexander Terekhov <terekhov@web.de> wrote:

>
> Joseph Seigh wrote:
> [...]
>> I'd like to see how you propose that the creation code manages
>> to see that 666 has been freed up for reuse and still sees 666
>> in memory as the lock owner.
>
> Same thing as with condvar queues.
>
> http://www.google.de/groups?selm=3C615A7E.C61764C7%40web.de
>

Any i/o or direct processor signaling also.  But that's extremely
contrived.  It would be a great deal of work to create an mechanism
that didn't use shared memory simply for the purpose of trying to
break programs using thread ids that way.  Plus, can you imagine
what would happen to your career at Microsoft if you tried to trip
up Richter that way? :)

I think it's more efficient to put the fetch of the thread id ahead
of any interlocked instructions on an ia32 processor.  You could make
it a speculative fetch and recheck it once you've incremented the count
if you cared.

-- 
Joe Seigh

.