The problem was to write a program which automatically solves Boulder Dash efficiently (the less turn you consume, the better).
As usual, I joined the lightning division (the first 24 hours of this 72 hours long contest) with C++. However, I decided to do something different this year because I was getting a bit bored of this kind of topcoder marathon match style problems. So, I wrote the boulder dash solver in Befunge:
http://shinh.skr.jp/dat_dir/bd.b98
I guess this is one of the biggest hand-written Befunge code. Technically, this isn't Befunge-93 code because it uses address space bigger than 80x25, so this code requires Funge-98 implementation like cfunge. However, I think it would be OK to claim this is a Befunge code because I only used Befunge-93 operations. Anyway, it wasn't easy to write this code, although I think I'm fairly good at Befunge (I've ever won a Befunge contest at codeforces). Like other participants, I didn't do almost nothing except for writing this code, but it was just 6 hours before the end of contest when I finished my simulator in Befunge...
The following is the full package I submitted. This includes a Befunge interpreter implementation written by me during this contest, some Makefile and README stuff, code for lightning division, etc.
http://shinh.skr.jp/dat_dir/icfp12.tgz
This is an attempt to describe something about my code. You can see a visualized image of dungeon around the bottom right of this image. This is just a memory dump. Befunge has 2D address space so we need no external visualizer.
BTW, I forgot to write something about ICFP programming contest 2011. It was great. The problem was better than all other contest problems I've ever seen and the organizers prepared a fairly stable duel server where participants can fight each other. I took 6th place. It seemed the 5th place team was also 1 person team and he got the judge's prize because of he was a one person team. So, I was the 2nd place one person team, yay.
