Back to the future!

Back to the future!

[Ludovic Courtès]

Hello!

There’s a brand new (ice-9 futures) module implementing futures:

  http://git.sv.gnu.org/cgit/guile.git/commit/?id=0d4e6ca38f1c51f5f92effc7d97c8b69eb85d071

(There used to be an implementation of futures in libguile, but it was
never included in a release because of various bugs.)

It basically does the job, though there are a couple of minor FIXMEs or
TODOs.

The implementation was inspired by that of libgomp, the GNU OpenMP
run-time support library.  It uses a pool of threads, normally one per
CPU core minus one, and offloads expression evaluation to them.  On a
single-core machine, you get an empty thread pool, meaning that futures
are evaluated sequentially, upon ‘touch’.

Feedback welcome!  :-)

Thanks,
Ludo’.

Re: Back to the future!

[Andy Wingo]

On Fri 03 Dec 2010 00:55, ludo@gnu.org (Ludovic Courtès) writes:

> There’s a brand new (ice-9 futures) module implementing futures:

Cool!

Andy
-- 
http://wingolog.org/

Re: Back to the future!

[Neil Jerram]

ludo@gnu.org (Ludovic Courtès) writes:

> Hello!
>
> There’s a brand new (ice-9 futures) module implementing futures:

Cool.  One thing I noticed:

>   (set! %futures (cons future %futures)) ;; FIXME: use a FIFO

I believe (ice-9 q) provides FIFOs, so you could use that.

      Neil

Re: Back to the future!

[Ludovic Courtès]

Hello!

Neil Jerram <neil@ossau.uklinux.net> writes:

> ludo@gnu.org (Ludovic Courtès) writes:

[...]

>>   (set! %futures (cons future %futures)) ;; FIXME: use a FIFO
>
> I believe (ice-9 q) provides FIFOs, so you could use that.

Right.  I’m not fond of its API, but we’ll see.

BTW, you might want to contemplate the restrictions of futures as
implemented by one of our competitor:

  http://docs.racket-lang.org/guide/performance.html#%28part._effective-futures%29

Both ‘*’ and ‘/’ can be used in Guile futures!  ;-)

Ludo’.

Re: Back to the future!

[Ludovic Courtès]

Hello!

I rebased ‘parallel’, ‘par-map’, and ‘par-for-each’ on top of futures
(commit c21a5ddcaf2b17d65355c977f37ee50375961d17).  The rationale is
documented in the same commit, with a recommendation against the ‘n-’
variants of these procedures.

Feedback welcome!

Ludo’.