Path: utzoo!utgpu!water!watmath!clyde!rutgers!cmcl2!nrl-cmf!ames!pasteur!ucbvax!hplabs!otter!kers
From: kers@otter.HP.COM (Christopher Dollin)
Newsgroups: comp.lang.lisp
Subject: Re: Composition/Partapply in CL Question
Message-ID: <1350004@otter.HP.COM>
Date: 13 Jan 88 16:03:10 GMT
References: <1350002@otter.HP.COM>
Organization: Hewlett-Packard Laboratories,Bristol,UK.
Lines: 21

One problem with using lexical variables freely (as in compose in the basenote)
is that although it may cons only a few words, the resulting closure may 
contain pointers to unused but arbitrarily large structures, thus permitting
garbage to be retained way part its normal collection. I'm afraid I don't
know enough about Lisp implementation to know if this is likely to be a 
problem: how is the environment usually represented?

[In the case of Pop11, lexical closures are implemented as concealed partial
applications, like lambda-lifting; hence no extra garbage is retained, but
the closure is more expensive to compute. It conses store by something like

    fixed overhead + 2 * number of free variables

where the fixed overhead is that for a simple procedure plus call to the
part-applyed procedure, and each free variable contributes some store for
the indirection headers and some for the instruction to push it as an argument.
If I have messed up the description I apologise in advance to John Gibson
and the rest of the Poplog team!]

Regards,
Kers.                               | "Why Lisp if you can talk Poperly?"