Re: testing a new Clisp Windows installer

[Pascal J. B. <pj...@in...>]

Wolfgang Dautermann <wol...@da...> writes:

> Hi all,
> I would also be interested in testing a new Clisp Windows (pre)release.
>
> I am a member of the Maxima-Team (Maxima is a CAS written in Common
> Lisp) and my main contribution is, that I create (crosscompiled)
> Windows-Installers for Maxima (for the 'crosscompiliation' for Lisp
> based programs I use Wine, which works well).
>
> I did start with Clisp as supported Lisp, now I also support SBCL. Clisp
> is slower, but - in my opinon - more hassle-free, it is still the
> default Lisp in my installers.
>
> So I am *certainly* interested in testing a new Clisp release for
> Windows - if it does work with my crosscompilation procedure. (If
> possible, both a 32 and a 64 bit version). [For the creation of my
> installer just a ZIP-File is easy to use, that was one of the reasons, I
> did start with Clisp (the other reason was, that it worked well with Wine)].
>
>
> Some more questions:
> Clisp is much slower than other Lisp implementations. On the Maxima
> website http://maxima.sourceforge.net/lisp.html
> there are some notes about the reasons (I am not sure, if they are correct):
>
> "Clisp is compiled to bytecodes, so Maxima running on Clisp is
> substantially slower than on Lisps compiled to machine instructions.
> Clisp computes floating-point operations in software, so floating-point
> operations in Clisp are much slower than in Lisp implementations that
> use hardware instructions for floating-point operations. On the other
> hand, Clisp makes use of the GMP library for arbitrary-precision integer
> computations."
>
> (a) Is "computes floating-point operations in software" correct?
> Nowadays? Can one disable that? I assume most processors can do it in
> hardware now.
> (b) "Clisp makes use of the GMP library" - how can one enable it? I
> compiled Clisp now from Daniels sourcecode, but libgmp seems not be used
> (checked using 'ldd'). And there is no configure-option '--use-libgmp'
> or something similar.

See Joerg's answers.

However, clisp has long-float numbers that can have arbitrary
precision, and therefore are implemented in software.

Does maxima use long-float and if yes, does it use
(setf ext:long-float-digits) to set the wanted precision (possibly big)?

If it only uses single-float and double-float, then it's done in
hardware.

There are also short-float numbers; I don't know how they're
implemented.  Probably they use single-float hardware and truncate the
mantissa on storage.


-- 
__Pascal Bourguignon__                 http://www.informatimago.com/
“The factory of the future will have only two employees, a man and a
dog. The man will be there to feed the dog. The dog will be there to
keep the man from touching the equipment.” -- Carl Bass CEO Autodesk