ok (04/13/83)
Some time ago there was an article in this newsgroup which said that
Guttierez' paper "Prolog Compared with Lisp" in the 1982 Lisp &
Functional Programming conference had PROVED that Prolog was GENERALLY
twice as slow as Lisp. Now that we are finally on USENET, I can reply
to that.
The title of this article is the title of an article that SigPlan
Notices have accepted which critically examines Guttierez' paper. What
his paper actually shows is that
a badly written Lisp program outperforms
a Prolog program which *deliberately* violates many aspects of
good Prolog style
This is hardly surprising.
When evaluating David Warren's claims about the relative efficiency
of Prolog and Lisp, or when evaluating my SigPlan article, do bear in
mind that they are comparisons between ONE version of Prolog and ONE
version of Lisp. There are faster versions of Lisp around, and most
other Prolog systems are just interpreters. Our claims about Prolog are
just that it is not INHERENTLY worse than Lisp for the tasks to which
both languages are suited.
By the way, Alan Mycroft and I have designed, and he has
implemented, a polymorphic type-checker for Prolog, based on Milner's
work for ML. It is only a couple of pages long. Interested, anyone?