Newsgroups: comp.lang.scheme
Path: cantaloupe.srv.cs.cmu.edu!das-news.harvard.edu!news2.near.net!MathWorks.Com!udel!princeton!nimaster.princeton.edu!blume
From: blume@beth.cs.princeton.edu (Matthias Blume)
Subject: Re: The expense of call/cc (was R4RS)
In-Reply-To: smcl@sytex.com's message of Wed, 5 Oct 1994 21:03:58 GMT
Message-ID: <BLUME.94Oct5221900@beth.cs.princeton.edu>
Originator: news@nimaster
Sender: news@Princeton.EDU (USENET News System)
Nntp-Posting-Host: beth.cs.princeton.edu
Organization: Princeton University
References: <36ut56$835@larry.rice.edu> <B2qqTc1w165w@sytex.com>
Date: Thu, 6 Oct 1994 02:18:59 GMT
Lines: 28

In article <B2qqTc1w165w@sytex.com> smcl@sytex.com (Scott McLoughlin) writes:

   drichter@owlnet.rice.edu (David Richter) writes:
   > the same expense, but popping a frame always costs at least one
   > instruction (decrement stack pointer) while garbage-collecting that
   > frame (using a copying garbage collector) is (asymptotically) free.
						  ----------------
   Howdy,
	   There's the rub.

Before posting hot-shot assertions like this I would recommend to do
some kind of analysis of one's own to be able argue on solid grounds.

I happen to know that Zhong Shao and Andrew Appel were *convinced*
(just like many are now) that the stack would beat the heap easily.
They went, implemented both, measured, and were very surprised that
the results didn't show any appreciatable differences.  *Then* they
went back, analysed the bahavior and tried to come up with some solid
*explanation* for this (as opposed to just making an unfounded *claim*
in order to pseudo-prove their pet-theory).

Of course, the collection isn't really free, but neither is the pop
off the stack.  But there is much more to be considered -- read the papers!

Well, let's rest this now

--
-Matthias
