SLUG uses huffman which simply cant be faster than a byte or word-coder with some bit-flags. If Id replace the huffman-coder with a byte-coder, SLUG would become quite a bit faster in most situations, but this would go along with a loss in ratio (as always, depending on the data).
Additionally, I compiled quick0 with GCC and compared its performance to the provided executable by Lasse. His compiler seems to create faster code than GCC (which I use), too.
So, to sum it up in 3 words: huffman and compiler.
