Do you referring to tasks like Sort N numbers using NxN processors? We did those task in Parallel algorithm class way back on college, using PRAM assembler.

I am not sure perl is really best environment for such play, since its threads are quite heavy-weight and have quite a cost for shared data (which are essential here). But for experiments you can build number of threads (like for each digit) and wait until less-significant digit thread provide its carry.

EDIT: You can also use library like Coro for lighter cooperative threads.

-- Roman