Date: Sun, 2 Apr 1995 10:08:45 -0400 From: "Paul F. Werkowski" <pw@snoopy.MV.COM> To: spaz@u.washington.edu Cc: hackers@FreeBSD.org Subject: Re: CLISP clarification, Was: New Snapshot...Good and Bad.... Message-ID: <199504021408.KAA00621@snoopy.mv.com> In-Reply-To: <Pine.OSF.3.91a.950401183626.15498B@saul2.u.washington.edu> (message from John Utz on Sat, 1 Apr 1995 19:11:49 -0800 (PST))
next in thread | previous in thread | raw e-mail | index | archive | help
>>>>> "John" == John Utz <spaz@u.washington.edu> writes:
John> Hokay; Allow me to clear up all of this confusion that i
John> started...
<...>
John> @ncftp>more ANNOUNCE @This is CLISP, a Common Lisp
John> implementation.
John> @ CLISP is mostly CLtL1 compliant, with some CLtL2
John> additions, including a @ CLOS subset. Many features of CLtL2
John> or dpANS CL are currently not @ supported.
John> @ The newest versions will always be available via anonymous
John> ftp from @ ma2s2.mathematik.uni-karlsruhe.de [129.13.115.2],
John> directory @ /pub/lisp/clisp/. @ Another ftp site carrying
John> CLISP is @ ftp.cs.cmu.edu [128.2.206.173], directory
John> user/ai/lang/lisp/impl/clisp/. ^^^^^^^^^^^^ Given the
John> thread of this conversation I think this mirror is kind of
John> significant. It would seem to me that if cmu wants to mirror
John> this particular implementation that would indicate that they
John> felt that it was a reasonable job that implements some
John> significant feature not found in there original.
Actually, The CMU site is a repository for lots of AI stuff,
practically anything that ever made it out of the labs, I think.
John> I would be willing to bet that significant feature is
John> SIZE!!!
John> @ Our Common Lisp CLISP @ * needs only 1.5 MB of memory
John> ^^^^^^^^^^^^^^^^
Hmm, that beats GCL which starts life at 2.4 MB. Seems like
it used to be smaller than that. Anyhow bear in mind that
Lisp grows like a wart once it has to actually do anything.
John> Did i read in someone else's posting that CMU CLISP
John> wants 16M?!!!
Anybodys Lisp wants lots of memory. GCL starts at 2.4 MB with
basic features. Add CLOS, LOOP, Conditions, etc to get some
of the CLtL2 functionality and it gets to 6.3 MB real quick.
Toss in CLX for X support and we are pushing 8MB. Add in
some application code, maybe a blackboard or rules system
and we start to exceed the VM limits on a 16 MB box. From
my experience, 16MB should be considered a minimal system
for running Lisp. 32MB is pretty nice. I saw one post where
a guy was running 192MB and was pretty happy.
John> @ * implements 99% of the CLtL1 standard, as well as some
John> extensions @ * can call your preferred editor @ * is freely
John> distributable
John> @ Get it via anonymous ftp from
John> ma2s2.mathematik.uni-karlsruhe.de @ [129.13.115.2], @
John> directory /pub/lisp/clisp/, or contact @ Bruno Haible
John> <haible@ma2s2.mathematik.uni-karlsruhe.de>.
John> If you are not hot and sweaty about getting the exact
John> particular package then i would seriously look at this. It
John> made right the heck up on 2.0R ( excepting __const char
John> *__const sys_errlist, of course ). It might be an improvment
John> to *have* this as opposed to *wondering* about the
John> portability of the full on CMU version.
Yeh, I might just check it out and see how it compares to GCL.
The neat feature of CMU Lisp
is that it is close to the new ANSI spec and has a native mode
compiler rather than having to go the Lisp-to-C route. The
bad news is that the i386 instruction set is not one of
alpha, Sparc or SGI. Also, it turns out that there is assembly
language hooks to the kernel for each implementation and one
needs an operational CMU-Lisp just to build the .h file for
the underlying C code. Looks like a challenge.
John> As i said originally, it flops on the current snap. It
John> compiles fine but seems to run out of room on my 8 meg 386
John> during the bytecompile stage ( this is a total wag , i infer
John> this because the lisp environment gets created from the c
John> code, executes, goes to work on a mess of *.lsp and dies in
John> one called clos.lsp, but what i don't know about lisp defies
Ah, but once you get to know it - you will either love it or
hate it. Close to the being the worlds oldest computer language
and one considered by many to be the worlds greatest systems
programming language of all time!
John> One of our vm gurus, i think it was DG but it might
John> have been Bruce, mentioned something about needing snapshots
John> to test out the current state of the vm code..so i would
John> think that this might be an elliptic reference to the grief
John> i was having.
Building Lisp on a small box WILL beat the cr ^H^H^H^H
mm, stress test any VM system.
John> in case you are wondering why i am torturing myself
John> with a language i can't comprehend,,, lisp is the lang of
John> choice for symbolic computation and the guy that wrote this
John> clisp has also done a gorgeous port of some classic symbolic
John> computation code. Given the fact that i dont even know what
John> the equivalent of printf is in lisp, it seemed logical to
(format t "The number is ~d~%" 1)
John> get his *environment* if i wanted to use his *port* with any
John> chance of having a successfull compilation. Which, btw,
John> turned out to be the case on 2.0R.
John> obviously, i eagerly await the next snap, but i think
John> i am stuck now until 2.1 because i dont see where i will
John> have time to do this again until the quarter is over...it
John> tooks three days of near continuous effort ( thank heaven my
John> wife was out of town... :-) )
The GCL port ought to work fine on -current.
Good luck.
Paul
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199504021408.KAA00621>