P. Norvig, in his Paradigms of AI Programming (1992) mentioned that good
Prolog systems were capable of 10,000 - 100,000 logical inferences per
second, or LIPS.
Two versions of Prolog in Lisp are developed in the book: interpreted and
compiled. The compiled version is 10-100 times faster, but it is still
benchmarked to do just 740 LIPS.
13 years since then, I'm sure all version will do much better, but the
relative difference in performance begs the question: Is there something
algorithmically naive about PAIP Prolog? How is it different from those
Prologs that run 100 times faster?
What is the best Prolog-in-CommonLisp today (I want to both understand how
it works and see the source) ?
alex goldman <·····@spamm.er> writes:
> P. Norvig, in his Paradigms of AI Programming (1992) mentioned that good
> Prolog systems were capable of 10,000 - 100,000 logical inferences per
> second, or LIPS.
>
> Two versions of Prolog in Lisp are developed in the book: interpreted and
> compiled. The compiled version is 10-100 times faster, but it is still
> benchmarked to do just 740 LIPS.
>
> 13 years since then, I'm sure all version will do much better, but the
> relative difference in performance begs the question: Is there something
> algorithmically naive about PAIP Prolog? How is it different from those
> Prologs that run 100 times faster?
>
> What is the best Prolog-in-CommonLisp today (I want to both understand how
> it works and see the source) ?
Don't know about source availability ( ask! ) but see
http://www.franz.com/support/tech_corner/#prolog-071504
and click the links to see an implementation that is
clocked in excess of 1 MLIPS. It happens to be based on
Norvig's PAIP version, but hasbeen highly optimized, and I
think it competes well with pure prologs available.
--
Duane Rettig ·····@franz.com Franz Inc. http://www.franz.com/
555 12th St., Suite 1450 http://www.555citycenter.com/
Oakland, Ca. 94607 Phone: (510) 452-2000; Fax: (510) 452-0182
alex goldman <·····@spamm.er> writes:
> What is the best Prolog-in-CommonLisp today (I want to both understand how
> it works and see the source) ?
Dunno about source, but from what I know, the best is Xanalys's, I mean,
Lispworks's
(BTW, congrats, guys, for making a quality product that deserves its own
company)
KnowledgeWorks. I think it uses a CLOS-extensible WAM.
--
Rahul Jain
·····@nyct.net
Professional Software Developer, Amateur Quantum Mechanicist
alex goldman wrote:
> P. Norvig, in his Paradigms of AI Programming (1992) mentioned that good
> Prolog systems were capable of 10,000 - 100,000 logical inferences per
> second, or LIPS.
>
> Two versions of Prolog in Lisp are developed in the book: interpreted and
> compiled. The compiled version is 10-100 times faster, but it is still
> benchmarked to do just 740 LIPS.
>
> 13 years since then, I'm sure all version will do much better, but the
> relative difference in performance begs the question: Is there something
> algorithmically naive about PAIP Prolog? How is it different from those
> Prologs that run 100 times faster?
>
> What is the best Prolog-in-CommonLisp today (I want to both understand how
> it works and see the source) ?
>
I don't know about the 'best' implementation, but the book
_The_Implementation_of_Prolog_
(http://www.amazon.com/exec/obidos/tg/detail/-/0691087571/qid=1105061338/sr=1-1/ref=sr_1_1/002-8268480-2650430?v=glance&s=books)
uses Common Lisp as the implementation language. (Ignore the low rating
at Amazon made by some moron.)
Paul Graham's _On_Lisp_ (http://www.paulgraham.com/onlisp.html) also
implements (a basic) Prolog in Common Lisp.
David Sletten