Xref: utzoo comp.sys.mac:32637 comp.lang.lisp:1827
Path: utzoo!attcan!uunet!cs.utexas.edu!tut.cis.ohio-state.edu!rutgers!psuvax1!psuvm.bitnet!cunyvm!nyser!mstr!pwa-b!philabs!linus!mbunix!bds
From: bds@mbunix.mitre.org (Barry D Smith)
Newsgroups: comp.sys.mac,comp.lang.lisp
Subject: Re: Is LISP viable on the Mac?
Keywords: LISP Macintosh development
Message-ID: <54602@linus.UUCP>
Date: 25 May 89 15:59:41 GMT
References: <1989May22.195611.756@mntgfx.mentor.com>
Sender: news@linus.UUCP
Reply-To: bds@mbunix (Smith)
Distribution: na
Organization: The MITRE Corporation, Bedford, Mass.
Lines: 16

We've written a large, complex scheduling system on Symbolics machines
which we've ported to the Mac II running Allegro CL with a 19 in Sony
Trinitron monitor.  The overall performance was slightly slower than
the same program running on a Sun 3/60 with Lucid 3.0, and quite a bit
slower than the progrm running on a Symbolics 3650.  Allegro's memory
management is pretty poor, and garbage collection was continously
triggered by the color graphics.  The reason for this was that all
graphics objects such as pixmaps reside on the Mac heap.  Allegro
cannot do a gc on them.  Since Allegro insists on keeping in its heap
as much of the system memory as possible, everytime a large graphics
object is created, Allegro must downsize its heap to make room, and
this triggers a global gc.  Similarly, whenever a large graphics
object is released, Allegro immediately increases its heap, which also
triggers a global gc.  In all, I think the poor memory management more
than anything caused us to give up the Mac effort.  We're now working
on both Symbolics machines and Sun 4's.