[Haskell-cafe] Deciding equality of functions.

I guess that deciding whether two functions are equal in most cases is algorithmically impossible. However maybe there exists quite a large domain of decidable cases? If so, how can I employ that in Haskell? It is a common situation when one has two implementations of the same function, one being straightforward but slow, and the other being fast but complex. It would be nice to be able to check if these two versions are equal to catch bugs in the more complex implementation.