[oggonachip] Re: pthread_cancel()

[Aaron J. Grier]

On Wed, Apr 17, 2002 at 12:29:34AM +0200, Pattara Kiatisevi wrote:
> On Thu, 14 Mar 2002, Aaron J. Grier wrote:
> 
> > On Thu, Mar 14, 2002 at 11:14:05AM +0100, Pattara Kiatisevi wrote:
> >
> > > Is there concept of rendezvous in classsic API? E.g. how would you
> > > implement this scenario? (It is not certain that task1 will connect to
> > > task2.meet() before or after task2 is waiting at task2.meet())
> > >
> > > Task 1:				Task2:
> > >
> > > <run whatever>			<run whatever>
> > >
> > > Connect task2.meet() ----------> Wait task2.meet()
> > > 				 <meeting point code, run while task1
> > > 				  blocks>
> > >
> > > <run whatever>			<run whatever>
> >
> > use a simple binary semaphore.
> 
> Hmm, but in case that task2 is faster than task1, then it released
> while noone was waiting (for obtaining). And then task1 blocks later,
> waiting for task2 to release whenever later. So it does not really
> correspond to the above rendezvous scheme?

you only show one task blocking in the above diagram.  :)

> (whoever who is faster will wait for another one and go together). Do
> I miss sth?

I think you can do it with two binary semaphores:

task1:			task2:

release(task1)		release(task2)
acquire(task2)		acquire(task1)

I think this does what you want.   it doesn't matter which task runs
first -- it will always wait for a release() from the other before it
continues running.

someone please correct me if I'm on the wrong track -- it's been too
long since my last OS class.  :)

-- 
  Aaron J. Grier  |   Frye Electronics, Tigard, OR   |  aaron at frye.com