Version française
Home     About     Download     Resources     Contact us    

This site is updated infrequently. For up-to-date information, please visit the new OCaml website at

Browse thread
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: 1997-04-16 (17:02)
From: Pawel Wojciechowski <Pawel.Wojciechowski@c...>

> You make it sound like the only point of threads is to take advantage of
> parallelism when the program is running on a multiprocessor machine.  
> A lot of people use threads for reasons that have nothing to do with
> (objective notions of) performance, for example in GUI programming.  

> -- 
> Frank Christoph                 Next Solution Co.      Tel: 0424-98-1811
>                             Fax: 0424-98-1500

Threads are very useful indeed! Even if they don't take advantage of 
multiprocessor shared memory architectures. I never questioned that. 
I'm sorry you misunderstood me. I just wanted to know why, e.g. the
architecure of Caml byte-code interpreter couldn't be multi-threaded. 
I'd like to thank Francois Rouaix (and others) who made it clear. We 
should wait for a truly concurrent memory management (garbage collector) 
to (o)Caml. As I understood the implementation is under way. I think 
having such an implementation ready to play with it would be fine.

In some applications, however, the "potential advantage" of threads 
(i.e. each thread executing on a separate processor), understood as 
*one* of many other advantages of threads, can be vital. A system 
which I'm implementing now is inherently concurrent. Ideally any 
scheduling decisions, as well as actions within the system, should 
be programmed in such a way that in a shared-memory multi-processor, 
parts of my system can run in true real-time parallel. A significant 
part of the project is implemented in (o)caml and just wondering 
whether I have to implement anything in C in order to improve QoS or 
stay with Caml for good.

> By your logic, there would seem to be no point in emulating concurrency
> on a sequential machine at all.

This is not my logic at all :) I know many examples where emulating 
concurrency on a sequenial machine proved to be very succesful. Perhaps 
one of the more spectacular examples would be an experimental language 
PICT implementing Robin Milner's concurrent Pi-calculus on a uniprocessor


   |   Pawel~ T Wojciechowski		  cambridge university    |
   |   hpage is       computer lab        |
   |   phome +44.1223 (3)34602                                    |