Developing the `Universal machine'

Hi! I've tried to solve very interesting task from the icfp. They propose to implement an interpreter of the byte-code for the specified `universal machine'. Well, I thought, such a program is a very easy to solve in C++, and what if I try to do it in Haskell? Said is done. But the result doesn't promise as much as I expected. First, the code is simple and clean indeed. Second, it works indeed. When I ran the benchmark by it, only simple tests passed, the other consumed too much memory. I wonder if there's a person who could help me out to tune the program...