[Icfp04-discuss] my Perl submission

[Mario Storti]

First of all, I have to congratulate the organizers for the choice of
problem and how the task was exposed. Very fun!!

I encouraged students from my class to participate. They sent at least
one submission (I think there are more). They were absolutely
delighted with the contest.

I have browsed the problems from other years and I think that this
one is certainly among the best if it is not the best!!! When I
suggested my students to participate I wondered whether it would be
interesting to participate or not. Now I'm very happy for having
encouraged them to participate. 

[In a previous post introducing me I inserted a link to the web page
of the course, but it had a typo. The correct page is
http://venus.ceride.gov.ar/aed mostly in spanish!! By the way, we live
in Santa Fe, Argentina. ]

My entry can be downloaded at 

http://venus.ceride.gov.ar/twiki/bin/view/Cimec/ICFPContestEntry

My web page is http://venus.ceride.gov.ar/mstorti

An extract of the features in my submission. 

* I wrote a server (server.pl) and compiler (zumb) in Perl. I planned
  first to make it in Scheme (Guile) but I am rather starting with
  Scheme and finally switched to Perl (which I'm using since long time
  ago).

* The server has a limited graphical output, but it can be called in
  background so that you can make matches in background in order to
  find the best ants. 

* The compiler has abstract labels, goto's, and subroutines (they are
  rather macros). After I wrote that I thought that perhaps this could
  be done better with M4, but since I'm more proficient in Perl I
  wrote the compiler in perl. The source files have extension .zb and
  you compile them as $ zumb ant.zb ant.as, where `ant.as' is the
  machine code ant. The compiler performed at 5secs every 1000 steps
  in a 100x100 map with 180ants (it is O(#ants), the size of the board
  is irrelevant for the speed). 

* Initially I wrote a very complex ant-brain that numbered the cells
  from 63 at the hill (home) decreasing towards infinity. These was
  very interesting, I had to write additions and substractions in
  binary with the compiler but in practice the ant turned out to be
  too slow.  These are ant2 and ant3 in the package, these were not
  submitted.

* After that (around sunday morning) I started writing an ant that
  marks towards away from the hill while searchng food and then
  unmarks while going home. Also it tries to find unmarked cells while
  searching food and marked cells while going home. These are ant4 and
  ant5. They used only one marker.

* Then I tried to figure out how to use the other markers. I have
  little time to experience, but I found beneficial to divide the ants
  in several species and let each species one marker. Then while
  searching food you look for any of the markers to be off. And going
  home any marker to be on. I can't how this works, but in matches
  this seems to improve.

* I never reached the level of martial arts. 

Happy hacking!!  

Mario