[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

What is GCL about? was Re: [Gcl-devel] numerical quirks

From: Camm Maguire
Subject: What is GCL about? was Re: [Gcl-devel] numerical quirks
Date: 11 Dec 2002 13:21:08 -0500

Greetings!  These and other similar discussions have prompted some
meditation on my part as to what GCL's focus or strategy should be --
what unique contribution does this project make; what justifies the
effort to keep it going?  Just having another free lisp alternative is
an inadequate answer, IMHO, with such high quality projects like CMUCL
and Clisp around.  Here is my thinking:

1) GCL is first about providing a high quality Common Lisp under the
   LGPL, thereby safeguarding user contributions to the project from
   possible third party proprietization for the indefinite future.
   Its official support by the FSF adds to the long-term reliability,
   availability and stability which is the intended purpose of the
   LGPL.  This distinguishes us from CMUCL, but not Clisp.

2) Secondly, GCL is about leveraging GCC technology to the maximum
   extent possible for the benefit of Lisp users, both in terms of
   portability and performance.  

3) Thirdly, GCL is about bringing existing open source lisp
   applications to as wide an audience as possible.  In other words,
   I'd hope people would find GCL 'practical', which in turn stems
   from its long history, and from lisp programmers designing their
   code to work with it over a long period of time.

If these are in general the three broad boundaries of our 'niche', then
what this means (to me) is that

1) We should import existing ansi lisp code/other high quality base
   lisp code wherever possible.  We're better off leveraging our
   currently limited human resources in our unique strengths rather
   than playing a late uphill catch-up game on the compliance front.
   This is *not* to say that compliance is not essential, --it
   absolutely is-- (and we're making encouraging progress!), but
   rather we should avoid reinventing the wheel and simply include
   relevant code from Clisp/CMUCL to achieve compliance in the most
   efficient manner possible.

2) Interfaces to external C libraries, faslink features, native object
   code loading, providing a frontend 'tree' output for the GCC
   backend, all become important.  As do portability and performance
   issues, both with GCL itself as well as with applications compiled
   with it.

I'd most like to hear any other thoughts!

By the way, although detailed performance analysis is still a ways
off, my first impression is that memory layout is the bottleneck on
modern systems.  Is anyone aware of algorithms to group conses onto
pages to minimize the number of links across page boundaries?  Sounds
like an NP complete traveling salesman problem.  In any case, this is
where image size still matters, IMHO, not in the lack of available
system memory.  I've committed a few patches to the maxima tree to
enable building with our ansi image, and it is significantly slower in
running 'make check'.  Likewise, Clisp, which produces an apparently
much smaller maxima image (rss size), performs about as fast as GCL on
'make check' even though it has no native code capability.  I hope we
can find time to look into these issues soon.

Take care, 

Peter Wood <address@hidden> writes:

> Hi,
> On Wed, Dec 11, 2002 at 12:48:51AM -0500, Raymond Toy wrote:
> > >(expt 2 (expt 2 (expt 2 (expt 2 2)))) causes SBCL and CMUCL to barf.
> > 
> > I think barf is too harsh.  They signal a continuable error saying the 
> > exponent is rather large and gives you the option of continuing the 
> > computation.
> > 
> Ok, I take 'barf' back.  I wasn't trying to start an argument.
> Fussing over stuff like (sin -1) is out of place at this stage, IMHO.
> Thats what I was trying to say.  We want sufficient bread and butter
> on the table before we can complain about the lack of champagne and
> caviar.
> Regards,
> Peter
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gcl-devel

Camm Maguire                                            address@hidden
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

reply via email to

[Prev in Thread] Current Thread [Next in Thread]