Re: Threads: monitors vs. primitives

From: William Kempf (sirwillard_at_[hidden])
Date: 2000-08-11 17:02:54

--- In boost_at_[hidden], "Greg Colvin" <gcolvin_at_u...> wrote:
> From: <scleary_at_j...>
> > > . . .  The problem here, though, is 
> > > correctness.  An event based implementation is subject to 
the "lost 
> > > wakeup bug".  See http://www.cs.wustl.edu/~schmidt/win32-cv-
1.html 
> > > for a good article on the issue.
> > 
> > That *is* a good article!  I stand corrected.  One question: 
should our
> > definition of CV include a signal() operation?
> > 
> > > BTW, I'm not new to the concept of a monitor.
> > 
> > Neither am I, of course.  :)
> 
> And neither is Proffessor Schmidt
>     http://www.cs.wustl.edu/~schmidt/PDF/monitor.pdf

Very true, and I've read this paper in the past.  Notice, however, 
that he's talking about a pattern, not a first class object.  He 
never advocates a first class monitor object over the mutex/condition 
primitives.  In fact, in his ACE library he includes the primitives 
and only gives you a sketchy template for a monitor.

I think I need to re-emphasise here.  I am _NOT_ against monitors.  I 
even think that a first class object for the concept may be 
appropriate here, though the pattern is easy enough to follow with 
just the primitives.  What I am against is removing the mutex in 
favor of such a monitor.

William Kempf