home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!olivea!apple!cambridge.apple.com!lynch@ils.nwu.edu
- From: lynch@ils.nwu.edu
- Newsgroups: comp.lang.lisp.mcl
- Subject: Re: Speed freak needs help
- Message-ID: <9212232033.AA14785@aristotle.ils.nwu.edu>
- Date: 23 Dec 92 20:33:51 GMT
- Sender: info-mcl-request@cambridge.apple.com
- Lines: 39
- Approved: comp.lang.lisp.mcl@Cambridge.Apple.C0M
-
-
- Maybe this is an insane question, but what sort of speed does Koza get with
- the Explorer?
- Is he getting 10-20 times yours?
- Or has he only used small test cases?
-
- ----------Quoted message follows:
-
- >Friends,
- >
- >I hate to say this, but we are now running into a speed problem with MCL and a
- >Quadra 950. It is of our own making, as I will explain, but I am urgently
- >seeking a means of running pretty standard Common Lisp at 10-20 times greater
- >speed than MCL 2.0 on a Quadra 950 36/400! Pretty GUIs are immaterial, and not
- >used - it is raw Lisp power that I seek (and we do accept that this will cost).
- >[Having said that, this does not mark a departure from MCL for 99% of our work,
- >and I am delighted to see that MCL is starting to get the recognition that it
- >deserves (congrats, MCL team!).]
- >
- >The reason is that we are starting to use John Koza's incredible Genetic
- >Programming (mail me if you have not heard of this, and I will send you a list
- >of refs. etc.) to optimise stack filters and other things which require
- >thousands of iterations of thousands of iterations of thousands of iterations.
- >Currently, one decent run takes over week on a Quadra 950, even after some
- >serious tuning of code. We are committed to using Lisp, and other languages
- >need not apply.
- >
- >One option that I am thinking of is using a farm of Quadras, but it would be
- >much neater if we could simply move our code onto another platform.
- >
- >John Koza extols TI Explorers, but after finally tracking down someone who
- >claims to have a small number of new Explorer systems, I am not sure whether we
- >would see anything like the necessary acceleration, and the systems are not
- >cheap. Suggestions as to whether an Explorer might suffice would be welcome
- >(and rough estimates on configuration would be even neater - given that the
- >code runs OK in 6 meg with MCL, but is clearly doing a lot of GC then).
- --
- --
- -- "TANSTAAFL" Rich lynch@ils.nwu.edu
-