Subject: Re: reducing number consing / cache experimental results
From: rpw3@rpw3.org (Rob Warnock)
Date: Thu, 25 Dec 2003 01:08:05 -0600
Newsgroups: comp.lang.lisp
Message-ID: <M5udnf25p89IEXeiRVn-uQ@speakeasy.net>
Richard Fateman  <rfateman@sbcglobal.net> wrote:
+---------------
| In my experiments you have to exceed the L2 cache size to see things
| slow down. 
+---------------

Ungar's original paper[1] proposing a fixed-location nursery for
generational GCs made exactly this point: For best performance
the size of the nursery should be somewhere around the size of
the secondary cache.


-Rob

[1] I don't have the reference at hand at the moment, but could dig
    it out of my hard-copy file of GC papers when I get back home
    next week.

-----
Rob Warnock			<rpw3@rpw3.org>
627 26th Avenue			<URL:http://rpw3.org/>
San Mateo, CA 94403		(650)572-2607