Home > Archive > PERL POE > November 2007 > Wide Finder
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
|
|
| Kevin Scaldeferri 2007-11-03, 10:36 pm |
| Okay, I gotta wonder, how come no one's submitted a Wide Finder
(http://www.tbray.org/ongoing/When/2...0/30/WF-Results)
implementation in POE (or any Perl for that matter)?
I spent a few minutes thinking about this, but the problem I had was
that I didn't know of a way to easily parallelize over multiple cores
that wasn't going to require me to write a ton of code. Are there
some POE modules that I couldn't find that do something like forking
a number of worker processes, then allowing me to send events between
them?
-kevin
| |
| David Davis 2007-11-03, 10:36 pm |
|
It looks like a competition.
He includes lines of code as a metric, which should tell you =20
something...
Well, anyway.
As Pierre, a commenter, suggested; you could write a script that runs =20=
Perl piped with xargs and the -n and -P options. No POE needed, and =20
it would be faster!
--=20
David Davis
=E2=80=A2 Software Engineer
=E2=80=A2 [ http://xant.us/ ]
On Nov 3, 2007, at 8:55, Kevin Scaldeferri <kevin@scaldeferri.com> =20
wrote:
> Okay, I gotta wonder, how come no one's submitted a Wide Finder =
(http://www.tbray.org/ongoing/When/2...0/WF-Results=20
> ) implementation in POE (or any Perl for that matter)?
>
> I spent a few minutes thinking about this, but the problem I had was =20=
> that I didn't know of a way to easily parallelize over multiple =20
> cores that wasn't going to require me to write a ton of code. Are =20
> there some POE modules that I couldn't find that do something like =20
> forking a number of worker processes, then allowing me to send =20
> events between them?
>
>
> -kevin
| |
| Chris Prather 2007-11-04, 4:51 am |
|
On Nov 3, 2007, at 9:39 p, David Davis wrote:
>
> It looks like a competition.
> He includes lines of code as a metric, which should tell you =20
> something...
>
> Well, anyway.
> As Pierre, a commenter, suggested; you could write a script that =20
> runs Perl piped with xargs and the -n and -P options. No POE =20
> needed, and it would be faster!
Well that's part of the issue, yes we have highly optimized single =20
threaded stuff but Tim Bray wants to take advantage of a 16 CPU =20
machine from sun. Distributed programming being the new paradigm.
[color=darkred]
> David Davis
> =95 Software Engineer
> =95 [ http://xant.us/ ]
>
> On Nov 3, 2007, at 8:55, Kevin Scaldeferri <kevin@scaldeferri.com> =20
> wrote:
>
[color=darkred]
Actually I've been using it to test some of my distributed code that =20
I've been working on in Moose and POE. I have a 130 line application =20
that wraps POE with Moose to distribute the processing across =20
multiple POE::Wheel::Run workers but the performance isn't high =20
enough for me to really want to trot it out to the world at large yet =20=
(~40s for ~4 million lines of input).
-Chris=
| |
| Kevin Scaldeferri 2007-11-04, 7:32 pm |
|
On Nov 3, 2007, at 7:39 PM, David Davis wrote:
> It looks like a competition.
It is, sort of. Mostly meant to be educational, though
> He includes lines of code as a metric, which should tell you
> something...
I think it's totally reasonable. It's interesting to see how
concisely different languages can express concurrency. Earlier in
the series of posts, he points out that in an ideal world, you'd like
it if you could express parallelization of the algorithm with just a
couple extra characters over the serial algorithm.
-kevin
|
|
|
|
|