Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site mit-eddie.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!barmar
From: barmar@mit-eddie.UUCP (Barry Margolin)
Newsgroups: net.lang,net.lang.lisp
Subject: Re: Using LISP for scientific programming? (gasp!)
Message-ID: <86@mit-eddie.UUCP>
Date: Tue, 15-Oct-85 00:48:21 EDT
Article-I.D.: mit-eddi.86
Posted: Tue Oct 15 00:48:21 1985
Date-Received: Wed, 16-Oct-85 05:29:00 EDT
References: <1057@sdcsvax.UUCP> <1418@umcp-cs.UUCP> <1071@sdcsvax.UUCP> <3785@garfield.UUCP>
Reply-To: barmar@mit-eddie.UUCP (Barry Margolin)
Distribution: net
Organization: MIT, Cambridge, MA
Lines: 32
Xref: watmath net.lang:1849 net.lang.lisp:576

In article <3785@garfield.UUCP> greg@hwcs.UUCP (Greg Michaelson) writes:
>LISP is slow because its memory model is non-linear and dynamic. FORTRAN
>has linear, static memory which corresponds 100% to VonN architecture.
>Most speed loss in LISP implementations, compiled or interpreted, is due to
>dynamic memory allocation and garbage collection.

I am a Lisp fan, but I mostly agree.  Lisp is inherently slow for some
things because some of its primitives are inherently expensive.

However, numeric, Fortranish applications often do not require much list
allocation.  And many Lisp implementations represent fixnum and flonum
objects as immediate data, rather than as objects allocated in the heap.
In these implementations, numeric applications will do mostly stack
allocations, not consing.  PDP-10 Maclisp, the one whose compiled code
was rated as good as Fortran, has special fixnum and flonum stacks which
are used for this optimization.  Numeric arrays are also very efficient
in most production Lisp implementations, especially if the arrays are
only allocated at the beginning of the run (as in Fortran, not
coincidentally).  The test which produced this comparison presumably
consisted of a Fortran program translated into Lisp, so it seems likely
that it didn't do much expensive consing.

My point is that both the inherent efficiency of a language (if this is
even a reasonable thing to discuss) and the efficiency of an
implementation depend heavily on the types of programs written using it.
Lisp programs are often slow because Lisp promotes a type of programming
that requires the use of dynamic memory allocation and garbage
collection.
-- 
    Barry Margolin
    ARPA: barmar@MIT-Multics
    UUCP: ..!genrad!mit-eddie!barmar