GC leak?

[Vladimir G. Ivanovic]

I went ahead and looked it up:

   http://www.parrotcode.org/docs/pdd/pdd25_concurrency.html

Note that this is a design document although there is some discussion 
of implementation. I don't know how much of the design has been 
implemented, but the file src/scheduler.c 
(http://svn.perl.org/viewvc/parrot/trunk/src/scheduler.c?view=log) 
does have code in it.

The JVM provides (if memory serves) just one construct to manage 
concurrency, the monitor, which uses locks on objects to do its job. 
Voila! Support for multithreading, etc. in one fell swoop since 
everything else is implemented in class libraries.

So, yes, concurrency is about data safety and liveliness, but the 
theory of the underlying implementation can be quite simple. Of 
course, the implications of the model (the practice) can take years to 
get completely correct, as in the Java case.

XEmacs on top of Parrot anyone? ;-)

--- Vladimir

P.S. In theory, practice and theory are the same. In practice, they 
are different.

on 02/03/2008 01:00 AM Stephen J. Turnbull said the following:
> Vladimir G. Ivanovic writes:
>  > on 02/02/2008 12:54 AM Stephen J. Turnbull said the following:
>  > > I've been following Python for
>  > > years and they still don't have an alternative to the Global
>  > > Interpreter Lock AFAIK.  
>  > 
>  > Does Parrot support multiple threads?
> 
> I don't know, but I would think any news that Parrot supports multiple
> threads well would set python-dev on fire, and I haven't noticed any
> firetrucks (except the ones heading toward San Diego :-( ) recently.
> 
> My understanding from the discussions on python-dev and from Simon
> Peyton-Jones's chapter "Beautiful Concurrency" in _Beautiful Code_ is
> that multithreading is not about the execution model (ie, the
> interpreter per se) as it is about the data model.  The problem is
> developing ways for the language to mark some data as safe, so that
> the interpreter doesn't need to lock them, while others are vulnerable
> to reentrancy etc and need to be protected.  Interpreter reentrancy
> per se is important, but it's a bounded task.  Fixing the language is
> not.
> 
> 


-- 

Vladimir G. Ivanovic


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: OpenPGP digital signature
Url : http://lists.xemacs.org/pipermail/xemacs-beta/attachments/20080203/8ed7bbf9/signature.pgp