[plt-scheme] Re: Using LibFFI along with threads

The problem with C libraries is that they do not always provide
callback scenarios, nor other ways to divide up functions in
smaller parts to get some 'time slicing'.

--Hans


Op 17/7/2007 schreef "Eli Barzilay" <eli at barzilay.org>:

>On Jul 17, Noel Welsh wrote:
>>  - It should be possible to specify an OS thread to use.  OS thread
>> creation is quite expensive on some systems.
>
>Wouldn't it make more sense to add threads when needed and keep some
>after they're done?  (Configurable like many popular servers.)  Or are
>you thinking of threads that hold some state?
>
>
>>  - The returned value should be a syncable value.  This is a simpler
>> programming model than the callback suggested in a previous thread.
>
>I think that they're pretty much equivalent in ease of use -- and more
>importantly, it's very easy to wrap a callback interface in a channel
>(simplified, untested):
>
>  ;; returns a channel, uses call/callback
>  (define (call/channel proc args)
>    (let ([c (make-channel)])
>      (call/callback proc args (lambda (res) (channel-put c res)))
>      c))
>
>  ;; invokes a callback when ready, uses call/channel
>  (define (call/callback proc args callback)
>    (thread (lambda () (callback (channel-get (call/channel proc args))))))
>
>Conclusion: C is a pain any way you look at it, so just go with
>whatever is easier -- and then add bells in Scheme.
>
>--
>          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
>                  http://www.barzilay.org/                 Maze is Life!
>
>