Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site g.cs.cmu.edu
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!harvard!seismo!rochester!pt.cs.cmu.edu!g.cs.cmu.edu!unilog
From: unilog@g.cs.cmu.edu (Unilogic)
Newsgroups: net.lang.lisp
Subject: Re: Re: Using LISP for scientific programming? (gasp!)
Message-ID: <319@g.cs.cmu.edu>
Date: Thu, 24-Oct-85 00:35:11 EDT
Article-I.D.: g.319
Posted: Thu Oct 24 00:35:11 1985
Date-Received: Fri, 25-Oct-85 04:34:44 EDT
Organization: Carnegie-Mellon University, CS/RI
Lines: 20


Close, but not quite.  Limiting the discussion to Vaxen, which seems
reasonable given the comparisons to Franz:

PSL has 3 different representations for integers: INUMs, FIXNUMs, and BIGNUMs.
I will ignore BIGNUMs in this discussion as they are a seperate issue and are
a loadable option in PSL anyway.  FIXNUMs are indeed stored on the heap and
have a tag of xxx (PSL tags are 5 bits).  INUMs are not stored on the heap and
have two distinct tags, xxx (POSINT) and xxx (NEGINT).  Using this
representation, so long as you keep your integers in the range -2^27 to 2^27-1
there will be no use of the heap when performing arithmetic.  In fact, there
is a package which causes the compiler to use in-line machine instructions
(ADDL, SUBL, etc.) if you are "sure" that you will not need any integers >
27 bits.  Every math function must check its result and if it is outside of
INUM range, allocate a longword on the heap and return a tagged address.

						David Gentzel
----------------
ARPA:	unilog@cmu-cs-g
UUCP:	seismo!cmu-cs-g!unilog