Re: [Caml-list] Efficient OCaml multicore -- roadmap?

[Gerd Stolpmann <info@gerd-stolpmann.de>]

Am Dienstag, den 12.04.2011, 23:36 -0400 schrieb Lucas Dixon:
> On 24/03/2011 19:13, Fabrice Le Fessant wrote:
> > On 03/24/2011 02:45 PM, Alexy Khrabrov wrote:
> >>> Where does the OCaml team stand on the multicore issues?  A year
> >>> or so ago, when there was a prototype parallel GC implementation,
> >>> IIRC, Xavier said it has to be done right.  So what are the
> >>> official plans and the status of integrating what volunteers had
> >>> done?
> >>>
> >>> WIth Scala having a robust actors model and AKKA kernel, and
> >>> Clojure built around efficient shared memory concurrency with
> >>> agents and references and STM, and Haskell also really parallel,
> >>> OCaml is lacking behind.  Furthermore, F# builds on strongly
> >>> parallel .NET, overcoming granddaddy.  With multicores common
> >>> even in laptops and iPads, we need an efficient  multicore OCaml!
> >>> Due to the model different from Haskell or Scala and Clojure, now
> >>> all on github, OCaml is both more stable and also is slower to
> >>> advance -- what do folks think about this situation?  How do you
> >>> do shared memory parallelism now?
> >>>
> >>> Cheers, Alexy
> >>>
> 
> The PolyML implementation of SML has had a multi-core runtime, without 
> slowing down single-threaded programs for a few years now.

How do you know that there are no slow-downs? Has it ever been tried to
get rid of all additional measures that are necessary to support
multicore? There might be e.g. architectural simplifications that would
lead to speed-ups for the single-threaded case.

>  It also has a 
> concurrent garbage collector. The main problem currently seems to be 
> that a lot of garbage gets produced, and memory access speeds end up 
> being the main bottle-neck. But we do see 4-5x speed-up on 8 core 
> machines for large real-world applications (in SML, our real world means 
> a big formal proof development, or graph rewriting at the moment... but 
> there's more fun stuff possible: http://kidkarolis.github.com/PolyChrome/)
> 
> Anyway, I suspect you'll run into some of the same issues as PolyML 
> encountered; there's a paper or two by David Matthews on this stuff; 
> (here's the first one I found: 
> http://portal.acm.org/citation.cfm?id=1708058&dl=ACM) and of course 
> there is the source code too:
> http://polyml.svn.sourceforge.net/viewvc/polyml/
> 
> Good luck with the OCaml version; it's been wanted for a long time! :)

You may be interested then in this pre-announcement I posted yesterday:

http://blog.camlcity.org/blog/multicore1.html

I was able to make good progress on my Netmulticore library, which -
without changing anything in the compiler or runtime - allows it to
write multicore-enabled shared memory programs following the
"multi-runtime" design (which is emulated by forking subprocesses).

More on this later. I'm working on a release.

Gerd
-- 
------------------------------------------------------------
Gerd Stolpmann, Bad Nauheimer Str.3, 64289 Darmstadt,Germany 
gerd@gerd-stolpmann.de          http://www.gerd-stolpmann.de
Phone: +49-6151-153855                  Fax: +49-6151-997714
------------------------------------------------------------