ICFP programming contest: results

ICFP programming contest: results

[Xavier Leroy]

Here are the official results for the ICFP 2000 programming contest:

First prize: the PLClub team from University of Pennsylvania
(J. Vouillon, H. Hosoya, V. Gapeyev, E. Sumii).  Language used: OCaml.

Second prize: the Camls 'R Us team from INRIA Rocquencourt
(S. Ailleret, P. Cuoq, D. Doligez, R. Harley, F. Le Fessant, X. Leroy,
A. Schmitt).  Language used: OCaml.

Third place: the Galois Connection team from the start-up company with
the same name (J. Launchbury and 4 other persons whose names I
couldn't write down).  Language used: Haskell.

Fourth place: a team of 13 distributed among Australia, Belgium and
the UK, including F. Henderson.  Language used: Mercury.

Judge's prize for the best sample image: a chess game by team Helikopter
(A. Rossberg, L. Kornstaed) from U. Saarlandes (I think).

There were 39 entries, with teams ranging from 1 to 13 persons.
Breakdown by language used:

        C, C++          7
        Clean           1
        Dylan           1
        Eiffel          1
        Haskell         6
        Java            6
        Mercury         1
        ML              9
        Perl            2
        Python          1
        Scheme          2
        Smalltalk       2

Not a bad year for OCaml.  And my personal congratulations to Jérôme
Vouillon and his friends for a truly outstanding performance!

- Xavier Leroy

Re: ICFP programming contest: results

[Brock]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: TEXT/PLAIN; charset=X-UNKNOWN, Size: 470 bytes --]




On Thu, 21 Sep 2000, Xavier Leroy wrote:

> ...
> Second prize: the Camls 'R Us team from INRIA Rocquencourt
> (S. Ailleret, P. Cuoq, D. Doligez, R. Harley, F. Le Fessant, X. Leroy,
> A. Schmitt).  Language used: OCaml.
> ...
> 
> Not a bad year for OCaml.  And my personal congratulations to Jérôme
> Vouillon and his friends for a truly outstanding performance!
> 
> - Xavier Leroy
> 

And to you at the Camls 'R Us team as well.

--Brock

Re: ICFP programming contest: results

[Eijiro Sumii]

Thanks for kind words, and sorry for having been silent so far -- we
thought it would be more fun to keep silence until the official
announcement, especially when people weren't aware of us.;)

I heard the 3rd place entry was 100 times faster than the 4th, and the
2nd (Xavier's) was 5.5 times faster than the 3rd, but the 1st (ours)
wasn't so much faster than the 2nd.  Doesn't this mean Caml is a
_much_ better language than others---though the contest judge seemed
somewhat reluctant to say Caml is "a fine programming tool for many
applications" and "the programming tool of choice for discriminating
hackers"---or, at least, its developers are much better programmers
than others?:-)  (Jerome was the main programmer in our team, and his
coding was amazingly fast.)

By the way, we are going to make a web page to describe what we did in
detail.  We'll announce it in this list when it's ready.

// Eijiro Sumii (http://www.yl.is.s.u-tokyo.ac.jp/~sumii/)
// 
// Ph.D. Student at Department of IS, University of Tokyo
// Visiting Scholar at Department of CIS, University of Pennsylvania

> From: Xavier Leroy <Xavier.Leroy@inria.fr>
> Subject: ICFP programming contest: results
> Date: Thu, 21 Sep 2000 09:12:56 +0200
> 
> Here are the official results for the ICFP 2000 programming contest:
...
> Not a bad year for OCaml.  And my personal congratulations to Jérôme
> Vouillon and his friends for a truly outstanding performance!
> 
> - Xavier Leroy

Re: ICFP programming contest: results

[Jean-Francois Monin]

Your web page will be of great interest.

Could you also tell us to what extent the knowledge of the
implementation of Ocaml was needed in order to get good performance ?

-- 
Jean-Francois Monin, France Telecom R&D DTL/MSV, Tel +33 2 96 05 26 79
2 av. Pierre Marzin, 22307 Lannion, France       Fax +33 2 96 05 39 45

-- 
Eijiro Sumii wrote :
 > Thanks for kind words, and sorry for having been silent so far -- we
 > thought it would be more fun to keep silence until the official
 > announcement, especially when people weren't aware of us.;)
 > 
 > I heard the 3rd place entry was 100 times faster than the 4th, and the
 > 2nd (Xavier's) was 5.5 times faster than the 3rd, but the 1st (ours)
 > wasn't so much faster than the 2nd.  Doesn't this mean Caml is a
 > _much_ better language than others---though the contest judge seemed
 > somewhat reluctant to say Caml is "a fine programming tool for many
 > applications" and "the programming tool of choice for discriminating
 > hackers"---or, at least, its developers are much better programmers
 > than others?:-)  (Jerome was the main programmer in our team, and his
 > coding was amazingly fast.)
 > 
 > By the way, we are going to make a web page to describe what we did in
 > detail.  We'll announce it in this list when it's ready.
 > 
 > // Eijiro Sumii (http://www.yl.is.s.u-tokyo.ac.jp/~sumii/)
 > // 
 > // Ph.D. Student at Department of IS, University of Tokyo
 > // Visiting Scholar at Department of CIS, University of Pennsylvania
 > 
 > > From: Xavier Leroy <Xavier.Leroy@inria.fr>
 > > Subject: ICFP programming contest: results
 > > Date: Thu, 21 Sep 2000 09:12:56 +0200
 > > 
 > > Here are the official results for the ICFP 2000 programming contest:
 > ...
 > > Not a bad year for OCaml.  And my personal congratulations to Jérôme
 > > Vouillon and his friends for a truly outstanding performance!
 > > 
 > > - Xavier Leroy

Re: ICFP programming contest: results

[Xavier Leroy]

> Could you also tell us to what extent the knowledge of the
> implementation of Ocaml was needed in order to get good performance ?

The only important thing to know was that a record with three float
fields is more efficient than a triple of floats.

Actually, all you needed to know (and much more) about the OCaml
performance model for the contest is documented in the OCaml
"Questions and answers" pages, especially
        http://caml.inria.fr/ocaml/speed.html
and
        http://caml.inria.fr/ocaml/numerical.html

- Xavier Leroy

WWW Page of Team PLClub (Re: ICFP programming contest: results)

[eijiro_sumii]

Dear Camlers,

> By the way, we are going to make a web page to describe what we did in
> detail.  We'll announce it in this list when it's ready.

Here it is.

  http://www.cis.upenn.edu/~sumii/icfp/

Enjoy,

Team PLClub

P.S.  Could you please add a link to the web page above in the contest
result page (http://www.cs.cornell.edu/icfp/contest_results.htm), Greg?

Re: WWW Page of Team PLClub (Re: ICFP programming contest: results)

[malc]

On Wed, 4 Oct 2000 eijiro_sumii@anet.ne.jp wrote:

> Dear Camlers,
> 
> > By the way, we are going to make a web page to describe what we did in
> > detail.  We'll announce it in this list when it's ready.
> 
> Here it is.
> 
>   http://www.cis.upenn.edu/~sumii/icfp/
> 
> Enjoy,
> 
> Team PLClub
> 
> P.S.  Could you please add a link to the web page above in the contest
> result page (http://www.cs.cornell.edu/icfp/contest_results.htm), Greg?
> 
> 
Duh! It cant even handle 1e-3! It enters endless loop on
http://caml.inria.fr/icfp00-contest/images/kal.gml, what a piece of ... :)
Just kidding, great work, rendered adrenalin at the speed i cant even
dream of.

But can someone of your team describe the rationale behind usage of arrays
as representation of matrices/vectors, likewise it will be also intersting
to hear corresponding rationale behind record based version of Caml's R Us
implementation.
 
-- 
mailto:malc@pulsesoft.com

Re: WWW Page of Team PLClub (Re: ICFP programming contest: results)

[Julian Assange]

> But can someone of your team describe the rationale behind usage of arrays
> as representation of matrices/vectors, likewise it will be also intersting
> to hear corresponding rationale behind record based version of Caml's R Us
> implementation.
>  
> -- 
> mailto:malc@pulsesoft.com

There's no use of Ocaml's objective features either. Jerome, can you comment
on that? The classes of 3d shapes were represented as types, for instance.
As the `O' in Ocaml, when do you think it is appropriate to use the objective
style? 

Cheers,
Julian.

Re: WWW Page of Team PLClub (Re: ICFP programming contest: results)

[eijiro_sumii]

> It cant even handle 1e-3!

We're sorry about that...:-)

> It enters endless loop on http://caml.inria.fr/icfp00-contest/images/kal.gml

It doesn't.  It's just slow for images like kal, because we forgot to
implement reflection cutting (as Camls 'R Us did).

Eijiro

Re: WWW Page of Team PLClub (Re: ICFP programming contest: results)

[eijiro_sumii]

P.S.

> But can someone of your team describe the rationale behind usage of arrays
> as representation of matrices/vectors,

(It was Jerome who decided that, but anyway...)  The reason was
convenience: we could write matrix/vector multiplication by using
for-loops, for instance.  As for the efficiency,
"records > tuples = arrays (without boundary checks)" and as for the
safety, "records = tuples > arrays", I suppose.

Re: WWW Page of Team PLClub (Re: ICFP programming contest: results)

[eijiro_sumii]

Oops, sorry,

> It doesn't.  It's just slow for images like kal, because we forgot to
> implement reflection cutting (as Camls 'R Us did).

I mean, Camls 'R Us did implement reflection cutting but we didn't.

Re: WWW Page of Team PLClub (Re: ICFP programming contest: results)

[eijiro_sumii]

> > (It was Jerome who decided that, but anyway...)  The reason was
> > convenience: we could write matrix/vector multiplication by using
> > for-loops, for instance.  As for the efficiency,
> > "records > tuples = arrays (without boundary checks)" and as for the
> > safety, "records = tuples > arrays", I suppose.

Oh, sorry, I meant "to represent matrices and vectors" only.

> Not exactly, for efficiency:
> 
> records = tuples = array without boundary checks

Excuse me, but I'm a bit confused here...  What did Xavier mean in the
message below then?

| Subject: Re: ICFP programming contest: results
| From: Xavier Leroy <Xavier.Leroy@inria.fr>
| Date: Tue, 26 Sep 2000 10:55:21 +0200
| 
| > Could you also tell us to what extent the knowledge of the
| > implementation of Ocaml was needed in order to get good performance ?
| 
| The only important thing to know was that a record with three float
| fields is more efficient than a triple of floats.

Re: WWW Page of Team PLClub (Re: ICFP programming contest: results)

[Pierre Weis]

> > Not exactly, for efficiency:
> > 
> > records = tuples = array without boundary checks
> 
> Excuse me, but I'm a bit confused here...  What did Xavier mean in the
> message below then?
> 
> | Subject: Re: ICFP programming contest: results
> | From: Xavier Leroy <Xavier.Leroy@inria.fr>
> | Date: Tue, 26 Sep 2000 10:55:21 +0200
> | 
> | > Could you also tell us to what extent the knowledge of the
> | > implementation of Ocaml was needed in order to get good performance ?
> | 
> | The only important thing to know was that a record with three float
> | fields is more efficient than a triple of floats.

Oups, sorry. There is a special treatment in the compiler for float
arrays and float records but no one for float tuples. So for the special
case of floats, array and record are more efficient than tuples. For
all other types there is no difference in efficiency.

Pierre Weis

INRIA, Projet Cristal, Pierre.Weis@inria.fr, http://cristal.inria.fr/~weis/

RE: ICFP programming contest: results

[Greg Morrisett]

As soon as we get back to Ithaca, we'll try to put up
more details about the entries.  I simply lumped the
ML dialects together because it fit well on the viewgraph --
I certainly didn't mean to make a political statement.  

I'd also like to offer my congratulations to both of
the O'Caml teams.  Their entries were truly spectacular.

-Greg

Re: ICFP programming contest: results

[Brian Rogoff]

On Fri, 22 Sep 2000, Xavier Leroy wrote:
> > In your language summary, was OCaml included among the ML category?
> 
> Several members of this list were surprised that the language summary
> only mentioned "ML" and not OCaml nor SML nor other ML variants.
> 
> This is how it appeared on the judge's presentation, but I take it
> that in their minds (as well as in mine), "ML" stands for all ML
> dialects: OCaml, Caml Light, SML, Moscow ML, etc.  (Similarly, they had
> only one line for C and C++, although both languages are definitely
> different.)
> 
> Actually, I like the use of "ML" as an umbrella for all these
> dialects, since it helps reduce the fragmentation of the
> F.P. community as viewed by outsiders.

Since we've entered the realm of marketing now, let me also suggest that
there are good reasons to identify your product as unique, and emphasize 
the distinctions with other (inferior :) products. It really depends on
the outsider who's view you want to influence; it is certainly easy to
hire an ML programmer to work on ML code, but maybe not so easy to port 
a large body of ML code to an ML compilation system. 

-- Brian

Re: ICFP programming contest: results

[Xavier Leroy]

> In your language summary, was OCaml included among the ML category?

Several members of this list were surprised that the language summary
only mentioned "ML" and not OCaml nor SML nor other ML variants.

This is how it appeared on the judge's presentation, but I take it
that in their minds (as well as in mine), "ML" stands for all ML
dialects: OCaml, Caml Light, SML, Moscow ML, etc.  (Similarly, they had
only one line for C and C++, although both languages are definitely
different.)

Actually, I like the use of "ML" as an umbrella for all these
dialects, since it helps reduce the fragmentation of the
F.P. community as viewed by outsiders.

- Xavier Leroy

Re: ICFP programming contest: results

[David McClain]

Outstanding!!!

In your language summary, was OCaml included among the ML category?

- DM

-----Original Message-----
From: Xavier Leroy <Xavier.Leroy@inria.fr>
To: caml-list@inria.fr <caml-list@inria.fr>
Date: Thursday, September 21, 2000 12:47 AM
Subject: ICFP programming contest: results


>Here are the official results for the ICFP 2000 programming contest:
>
>First prize: the PLClub team from University of Pennsylvania
>(J. Vouillon, H. Hosoya, V. Gapeyev, E. Sumii).  Language used: OCaml.
>
>Second prize: the Camls 'R Us team from INRIA Rocquencourt
>(S. Ailleret, P. Cuoq, D. Doligez, R. Harley, F. Le Fessant, X. Leroy,
>A. Schmitt).  Language used: OCaml.
>
>Third place: the Galois Connection team from the start-up company with
>the same name (J. Launchbury and 4 other persons whose names I
>couldn't write down).  Language used: Haskell.
>
>Fourth place: a team of 13 distributed among Australia, Belgium and
>the UK, including F. Henderson.  Language used: Mercury.
>
>Judge's prize for the best sample image: a chess game by team Helikopter
>(A. Rossberg, L. Kornstaed) from U. Saarlandes (I think).
>
>There were 39 entries, with teams ranging from 1 to 13 persons.
>Breakdown by language used:
>
>        C, C++          7
>        Clean           1
>        Dylan           1
>        Eiffel          1
>        Haskell         6
>        Java            6
>        Mercury         1
>        ML              9
>        Perl            2
>        Python          1
>        Scheme          2
>        Smalltalk       2
>
>Not a bad year for OCaml.  And my personal congratulations to Jérôme
>Vouillon and his friends for a truly outstanding performance!
>
>- Xavier Leroy
>