speed, etc.

Zuurbier, E. - AMSXE Erik.Zuurbier@KLM.NL
Thu, 28 Jan 1999 10:33:42 +0100


>What with the speedup of
>hardware, I dare say Clean now runs as fast as ANY compiled language
>did then.  In another few years, Clean will run as fast on Merced
>chips or whatever as C does now.

I think hardware improvement does not solve the problem completely.
I read an article "Structuring Depth-First Search Algorithms in Haskell"
by David J. King and John Launchbury in which they pointed out that
they had succeeded in turning a quadratic complexity into a linear one,
thanks to the use of monads. By the way, their monadic approach
can be converted to using uniqueness types in a very straightforward way,
also achieving linear complexity. As a side effect, imho the monadic
part changes from incomprehensible gibberish to beautiful prose :-).

Now if there remain other algorithms for which the time complexity is
better in an imperative language, hardware improvements will do
little to solve that. So yes, there may remain projects for which
FP's are too slow, and opportunities for researchers to speed them up.

Erik Zuurbier