Dylan Hackers entry won second prize at ICFP 2001 
Author Message
 Dylan Hackers entry won second prize at ICFP 2001

Hi!

I'm just back from Florence, the location of this year's
International Conference on Functional Programming:

  http://www.*-*-*.com/

At the conference, the winners of this year's ICFP Programming
Contest were announced, and I'm proud to say that we made the
second place:

  http://www.*-*-*.com/

and that it is now Official Scientific Truth[tm] that

  Dylan is a fine programming tool for many applications

The Dylan Hackers team consisted of (in alphabetical order)
Andreas Bogk, Chris Double, Bruce Hoult and Peter Housel.

Andreas



Tue, 24 Feb 2004 05:03:37 GMT  
 Dylan Hackers entry won second prize at ICFP 2001
Congratulations to all who made it!!!

Tomorrow

Quote:
>   Dylan is a fine programming tool for many applications

will display in big letters at my workplace!

Cool, guys!

    Gabor

----------

Quote:


>Subject: Dylan Hackers entry won second prize at ICFP 2001
>Date: Don, 6. Sep 2001 23:15 Uhr

> Hi!

> I'm just back from Florence, the location of this year's
> International Conference on Functional Programming:

>   http://cristal.inria.fr/ICFP2001/

> At the conference, the winners of this year's ICFP Programming
> Contest were announced, and I'm proud to say that we made the
> second place:

>   http://cristal.inria.fr/ICFP2001/prog-contest/

> and that it is now Official Scientific Truth[tm] that

>   Dylan is a fine programming tool for many applications

> The Dylan Hackers team consisted of (in alphabetical order)
> Andreas Bogk, Chris Double, Bruce Hoult and Peter Housel.

> Andreas



Tue, 24 Feb 2004 06:17:59 GMT  
 Dylan Hackers entry won second prize at ICFP 2001
Congratulations, and great work.  Hip-hip-hip hooray for Captains Spaulding!


Quote:
> Hi!

> I'm just back from Florence, the location of this year's
> International Conference on Functional Programming:

>   http://cristal.inria.fr/ICFP2001/

> At the conference, the winners of this year's ICFP Programming
> Contest were announced, and I'm proud to say that we made the
> second place:

>   http://cristal.inria.fr/ICFP2001/prog-contest/

> and that it is now Official Scientific Truth[tm] that

>   Dylan is a fine programming tool for many applications

> The Dylan Hackers team consisted of (in alphabetical order)
> Andreas Bogk, Chris Double, Bruce Hoult and Peter Housel.

> Andreas



Tue, 24 Feb 2004 10:29:12 GMT  
 Dylan Hackers entry won second prize at ICFP 2001
Interesting.  The guy Bruce thought would win didn't even make top 5.  A
bit of a surprise there since I think he had the other top 3 finishers
picked out. :-|
Quote:

> Hi!

> I'm just back from Florence, the location of this year's
> International Conference on Functional Programming:

>   http://cristal.inria.fr/ICFP2001/

> At the conference, the winners of this year's ICFP Programming
> Contest were announced, and I'm proud to say that we made the
> second place:

>   http://cristal.inria.fr/ICFP2001/prog-contest/

> and that it is now Official Scientific Truth[tm] that

>   Dylan is a fine programming tool for many applications

> The Dylan Hackers team consisted of (in alphabetical order)
> Andreas Bogk, Chris Double, Bruce Hoult and Peter Housel.

> Andreas



Tue, 24 Feb 2004 11:12:15 GMT  
 Dylan Hackers entry won second prize at ICFP 2001

Quote:

> it is now Official Scientific Truth[tm] that

>   Dylan is a fine programming tool for many applications

> The Dylan Hackers team consisted of (in alphabetical order)
> Andreas Bogk, Chris Double, Bruce Hoult and Peter Housel.

Well done all.

__Jason



Tue, 24 Feb 2004 17:05:33 GMT  
 Dylan Hackers entry won second prize at ICFP 2001

Quote:

> Hi!

> I'm just back from Florence, the location of this year's
> International Conference on Functional Programming:

>   http://cristal.inria.fr/ICFP2001/

> At the conference, the winners of this year's ICFP Programming
> Contest were announced, and I'm proud to say that we made the
> second place:

>   http://cristal.inria.fr/ICFP2001/prog-contest/

> and that it is now Official Scientific Truth[tm] that

>   Dylan is a fine programming tool for many applications

Thanks for attending for us, Andreas.

There is now a write-up at

   <http://www.hoult.org/~bruce/dylan-hackers.html>

-- Bruce



Tue, 24 Feb 2004 17:46:36 GMT  
 Dylan Hackers entry won second prize at ICFP 2001

Quote:

> At the conference, the winners of this year's ICFP Programming
> Contest were announced, and I'm proud to say that we made the
> second place

Congratulations! That's brilliant.

Quote:
> and that it is now Official Scientific Truth[tm] that

>   Dylan is a fine programming tool for many applications

The modesty of this claim underlines why Dylan is a better programming
language socially as well as technically. :-)

- Rob.



Tue, 24 Feb 2004 19:30:46 GMT  
 Dylan Hackers entry won second prize at ICFP 2001

Quote:

> Interesting.  The guy Bruce thought would win didn't even make top 5.  A
> bit of a surprise there since I think he had the other top 3 finishers
> picked out. :-|

This is easy to explain: Bruce looked at both execution speed and
optimizer quality, whereas the real judges didn't look at execution
speed at all.

And everybody who gets the joke in the name of the winning program,
which is "LALR(5000000)", must admit that they deserved the first
place.

Andreas

--
"In my eyes it is never a crime to steal knowledge. It is a good
theft. The pirate of knowledge is a good pirate."
                                                       (Michel Serres)



Tue, 24 Feb 2004 20:28:23 GMT  
 Dylan Hackers entry won second prize at ICFP 2001

Quote:

>At the conference, the winners of this year's ICFP Programming
>Contest were announced, and I'm proud to say that we made the
>second place:

>The Dylan Hackers team consisted of (in alphabetical order)
>Andreas Bogk, Chris Double, Bruce Hoult and Peter Housel.

Congratulations, guys!

Neel



Wed, 25 Feb 2004 06:38:57 GMT  
 Dylan Hackers entry won second prize at ICFP 2001

Quote:


> > Interesting.  The guy Bruce thought would win didn't even make top 5.  A
> > bit of a surprise there since I think he had the other top 3 finishers
> > picked out. :-|

> This is easy to explain: Bruce looked at both execution speed and
> optimizer quality, whereas the real judges didn't look at execution
> speed at all.

While that's perfectly true, the real reason is that his program
didn't manage its RAM use well enough on some of the later test files
and wen't into swap-yourself-to-death mode and failed to terminate
within the time limit.

Quote:
> And everybody who gets the joke in the name of the winning program,
> which is "LALR(5000000)", must admit that they deserved the first
> place

Yeah.  their write-up boggles me.

But, y'know what?  From what I've seen in all the new stuff the judges
posted today, If I'd only had the extra half-hour necessary to
implement dynamic adjustment of my "fudge factors" then we would have
beaten them solidly even with our lack of theoretical elegance.  Dylan
Hacker was just dragged down too much in the final rounds by a lot of
cases in which it ran out of time and had to fall back to the much
poorer Dylan Lightning result.  You can see this because it's the same
output size as for Dylan Lightning.  When the judges re-ran those
files with much longer timeouts Dylan Hacker kicked booty.

The Haskell Carrots clearly had the better algorithm, but I think we
had a big edge on them for sheer brute-force speed.

I'm pretty sure that speed is the reason that Andreas' "Beam Search"
didn't do better -- it was a good algorithm and elegantly-written, but
it hadn't been tweaked into producing efficient code.  With "Dylan
Hacker" I at least found time to grep the intermediate C code and find
any cases of gf-call in critical loops and do whatever was necessary
to nuke them.

Again with the later tests with much longer timeouts Beam Search
started doing much better, finding better solutions than Dylan Hacker
in several cases -- Dylan hacker was quite well tuned but had a fixed
level of effort and had no way to take advantage of any extra time
that was available.  Well, that was an explicit decision of mine so I
can't complain...

All said, we did great.  We submitted three quite different programs.
The worst of them (written in 24 hours) got 13th place, the two
72-hour entries both got into the final ten, and the best of them got
2nd.



Wed, 25 Feb 2004 10:10:27 GMT  
 Dylan Hackers entry won second prize at ICFP 2001

Quote:

> There is now a write-up at

>    <http://www.hoult.org/~bruce/dylan-hackers.html>

This page has been hit a 1427 times so far today.

The previous one (with the analysis of the first round results) got
about 450 hits in the four weeks since I posted it, and another 400 so
far today.

I guess that's a SlashDotting, but not a very extreme one :-)

-- Bruce



Wed, 25 Feb 2004 10:19:21 GMT  
 
 [ 11 post ] 

 Relevant Pages 

1. Sterling Prize Drawing - 11 prizes to be won by May 31

2. Speech - Turner Prize [ 10th Dec 2001 ]

3. call for papers ICFP 2001

4. call for papers ICFP 2001

5. ICFP Programming Contest 2001

6. call for papers ICFP 2001

7. ICFP 2001 contest - Python team, anyone?

8. call for papers ICFP 2001

9. APL-L Digest - 24 Sep 2001 to 25 Sep 2001 (#2001-223)

10. APL-L Digest - 14 Jun 2001 to 15 Jun 2001 (#2001-136)

11. APL-L Digest - 22 Mar 2001 to 23 Mar 2001 (#2001-63)

12. APL-L Digest - 13 Feb 2001 to 14 Feb 2001 (#2001-34)

 

 
Powered by phpBB® Forum Software