Multiprocessor support in OCaml

Multiprocessor support in OCaml

[Jason Ganetsky]

[-- Attachment #1: Type: text/plain, Size: 708 bytes --]

Hi all,

I'm new to this list, and new to OCaml (although, have some experience with
SML).

Anyway, I have recently written an OCaml thread pool implementation, on top
of the Thread and Event modules. I did this for the purpose of exploiting an
SMP system I have, and was a disappointed to read today that OCaml doesn't
support multiprocessor systems.

I played around with it a little, and discovered that by liberally calling
Thread.yield, I do cajole my threads into running on multiple processors. Is
this behavior normal, or have I discovered a problem with the Thread module?
I'm certainly happy that I can get it to use my SMP... but I will stop it at
once if you tell me that this is unsafe.

-Jason

[-- Attachment #2: Type: text/html, Size: 750 bytes --]

Re: [Caml-list] Multiprocessor support in OCaml

[Richard Jones]

On Sun, Apr 22, 2007 at 03:42:09AM -0400, Jason Ganetsky wrote:
> I'm new to this list, and new to OCaml (although, have some experience with
> SML).

There's a beginner's list:
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

> Anyway, I have recently written an OCaml thread pool implementation, on top
> of the Thread and Event modules. I did this for the purpose of exploiting an
> SMP system I have, and was a disappointed to read today that OCaml doesn't
> support multiprocessor systems.
> 
> I played around with it a little, and discovered that by liberally calling
> Thread.yield, I do cajole my threads into running on multiple processors. Is
> this behavior normal, or have I discovered a problem with the Thread module?
> I'm certainly happy that I can get it to use my SMP... but I will stop it at
> once if you tell me that this is unsafe.

The garbage collector doesn't support concurrency, so there's a big
global lock around all OCaml code.

http://caml.inria.fr/pub/ml-archives/caml-list/2002/11/64c14acb90cb14bedb2cacb73338fb15.en.html

Rich.

-- 
Richard Jones
Red Hat

Re: [Caml-list] Multiprocessor support in OCaml

[Xavier Leroy]

> Anyway, I have recently written an OCaml thread pool implementation, on
> top of the Thread and Event modules. I did this for the purpose of
> exploiting an SMP system I have, and was a disappointed to read today
> that OCaml doesn't support multiprocessor systems.

You are correct that OCaml *threads* do not exploit multiprocessing.
Basically, only one OCaml thread can run at a time.

You can still get parallelism in several ways.  First, external C
libraries called from OCaml can run in parallel with OCaml code
provided the OCaml/C interface for these libraries makes uses of the
"blocking section" mechanism.  Second, process-level parallelism works
very well with programs written in message-passing style, using e.g.
OcamlMPI or OCamlP3L.

> I played around with it a little, and discovered that by liberally
> calling Thread.yield, I do cajole my threads into running on multiple
> processors.

This is an illusion.  Thread.yield gives more opportunities to the OS
scheduler to reschedule a Caml thread on a different processor, but
you're not gaining parallelism this way and you might actually lose
performance (because of cache ping-pong effects and the like).

- Xavier Leroy

RE: [Caml-list] Multiprocessor support in OCaml

[Don Syme]

Just to mention there is a way of getting multiple concurrently executing OCaml threads in a program, which I discovered a while back: you can statically link multiple independent copies of the OCaml runtime, each into its own DLL (on Windows). This allows multiple independent OCaml threads to run concurrently.

I presume this technique works well enough for SMP up to 2-4 processors, though have never done any serious performance testing.

The OCaml programs must not, of course, trade OCaml values, but can communicate in-process by other means (e.g. shared C memory or some other message passing technique).

Regards,
Don

P.S. I've only used this technique on Windows.


-----Original Message-----
From: caml-list-bounces@yquem.inria.fr [mailto:caml-list-bounces@yquem.inria.fr] On Behalf Of Xavier Leroy
Sent: 22 April 2007 11:30
To: Jason Ganetsky
Cc: caml-list@yquem.inria.fr
Subject: Re: [Caml-list] Multiprocessor support in OCaml

> Anyway, I have recently written an OCaml thread pool implementation, on
> top of the Thread and Event modules. I did this for the purpose of
> exploiting an SMP system I have, and was a disappointed to read today
> that OCaml doesn't support multiprocessor systems.

You are correct that OCaml *threads* do not exploit multiprocessing.
Basically, only one OCaml thread can run at a time.

You can still get parallelism in several ways.  First, external C
libraries called from OCaml can run in parallel with OCaml code
provided the OCaml/C interface for these libraries makes uses of the
"blocking section" mechanism.  Second, process-level parallelism works
very well with programs written in message-passing style, using e.g.
OcamlMPI or OCamlP3L.

> I played around with it a little, and discovered that by liberally
> calling Thread.yield, I do cajole my threads into running on multiple
> processors.

This is an illusion.  Thread.yield gives more opportunities to the OS
scheduler to reschedule a Caml thread on a different processor, but
you're not gaining parallelism this way and you might actually lose
performance (because of cache ping-pong effects and the like).

- Xavier Leroy

_______________________________________________
Caml-list mailing list. Subscription management:
http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
Archives: http://caml.inria.fr
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
Bug reports: http://caml.inria.fr/bin/caml-bugs

Re: [Caml-list] Multiprocessor support in OCaml

[Erik de Castro Lopo]

Jason Ganetsky wrote:

> Hi all,
> 
> I'm new to this list, and new to OCaml (although, have some experience with
> SML).
> 
> Anyway, I have recently written an OCaml thread pool implementation, on top
> of the Thread and Event modules. I did this for the purpose of exploiting an
> SMP system I have, and was a disappointed to read today that OCaml doesn't
> support multiprocessor systems.

For real multi-procesor parallelism, have a look at this:

    http://www.pps.jussieu.fr/~dicosmo/ocamlp3l/

Haven't tried it myself, but its on my todo list.

Erik
-- 
+-----------------------------------------------------------+
  Erik de Castro Lopo
+-----------------------------------------------------------+
"life is too long to be an expert at harmful things, including 
such evilness as C++ and perl." -- Erik Naggum

Re: [Caml-list] Multiprocessor support in OCaml

[Jon Harrop]

On Sunday 22 April 2007 11:58, Erik de Castro Lopo wrote:
> For real multi-procesor parallelism, have a look at this:
>
>     http://www.pps.jussieu.fr/~dicosmo/ocamlp3l/
>
> Haven't tried it myself, but its on my todo list.

Has anyone tries this? Got any demos? I've been meaning to look at it for a 
while as well... :-)

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
The F#.NET Journal
http://www.ffconsultancy.com/products/fsharp_journal/?e

Re: Multiprocessor support in OCaml

[Zheng Li]

Hi,

I'm working on a process back-end of STM library. It's now supported by Google
SOC and expected to release after the summer (and maybe earlier). With it, you
will be able to do shared-memory (supposing that's the style your want)
parallel programming based on processes, which in turn gives you speedup. 

If interested, you can have a taste first through the (vm)thread back-end
currently available (check my homepage below), though it won't really speed up
your program because of the well-known global lock of OCaml threads.

"Jason Ganetsky" <jason.ganetsky@gmail.com> writes:
> Hi all,
> I'm new to this list, and new to OCaml (although, have some experience with
> SML).Anyway, I have recently written an OCaml thread pool implementation, on
> top of the Thread and Event modules. I did this for the purpose of exploiting
> an SMP system I have, and was a disappointed to read today that OCaml doesn't
> support multiprocessor systems.
>
> -Jason

-- 
Zheng Li
http://www.pps.jussieu.fr/~li

Re: [Caml-list] Re: Multiprocessor support in OCaml

[Jason Ganetsky]

[-- Attachment #1: Type: text/plain, Size: 1860 bytes --]

On 4/22/07, Jason Ganetsky <jason.ganetsky@gmail.com> wrote:
>
> Well, the solution I'm going for now is to load all my data up, call
> Gc.Compact(), and then fork off child processes. The workload that I'm
> parallelizing is read-only... so I think this will work well with Linux
> copy-on-write forking.
>
> On 4/22/07, Zheng Li <li@pps.jussieu.fr> wrote:
> >
> >
> > Hi,
> >
> > I'm working on a process back-end of STM library. It's now supported by
> > Google
> > SOC and expected to release after the summer (and maybe earlier). With
> > it, you
> > will be able to do shared-memory (supposing that's the style your want)
> > parallel programming based on processes, which in turn gives you
> > speedup.
> >
> > If interested, you can have a taste first through the (vm)thread
> > back-end
> > currently available (check my homepage below), though it won't really
> > speed up
> > your program because of the well-known global lock of OCaml threads.
> >
> > "Jason Ganetsky" <jason.ganetsky@gmail.com> writes:
> > > Hi all,
> > > I'm new to this list, and new to OCaml (although, have some experience
> > with
> > > SML).Anyway, I have recently written an OCaml thread pool
> > implementation, on
> > > top of the Thread and Event modules. I did this for the purpose of
> > exploiting
> > > an SMP system I have, and was a disappointed to read today that OCaml
> > doesn't
> > > support multiprocessor systems.
> > >
> > > -Jason
> >
> > --
> > Zheng Li
> > http://www.pps.jussieu.fr/~li <http://www.pps.jussieu.fr/%7Eli>
> >
> > _______________________________________________
> > Caml-list mailing list. Subscription management:
> > http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
> > Archives: http://caml.inria.fr
> > Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> > Bug reports: http://caml.inria.fr/bin/caml-bugs
> >
>
>

[-- Attachment #2: Type: text/html, Size: 3249 bytes --]

Re: [Caml-list] Re: Multiprocessor support in OCaml

[Richard Jones]

On 4/22/07, Jason Ganetsky <jason.ganetsky@gmail.com> wrote:
>
>Well, the solution I'm going for now is to load all my data up, call
>Gc.Compact(), and then fork off child processes. The workload that I'm
>parallelizing is read-only... so I think this will work well with Linux
>copy-on-write forking.

You might also want to look at the Ancient library
(http://merjis.com/developers/ancient) which will allow you to share
data read-only between unrelated processes, backed by a file.

Rich.

-- 
Richard Jones
Red Hat