Question regarding Sbcl



> Seems to me you should be asking on the sbcl list instead of maxima.  IIRC
> clisp cannot build sbcl, but cmucl or ccl can.
>

sure, but the main question was (below) whether it might be worth the effort
at all (w.r.t. Maxima).
And I also thought there might be some general interest in these issues.
>From my point of view, the situation using Lisp seems pretty bad (see below).
 
> Or just go grab a binary of sbcl (or cmucl or ccl).
> 

I do not use binaries for various reasons (taking them altogether):
1. All we are doing is going to be redistributed, and we want to provide
a complete and (nearly) automated build.
2. It seems to me essential for the spirit of open-source to be able to
build from sources (what is the source-code good for otherwise?).
3. I do not trust a system without the sources (working sources that is).
4. I do not trust the quality of a system which can not be built from
sources (I think one should make an effort like Sage, or actually also
our "library", to provide everything that it needed in a "standard Linux
environment").

If the bigger Linux distributions have it, also fine with us.

Now from this point of view only CLisp is fully available --- it
is available in the larger Linux distributions, and it builds from
sources.
Next comes Ecl, which can be built from sources.
And that seems to be it:

 - Cmucl requires a binary (and at least for Suse it doesn't seem to be
   available).
 - CCL again requires a binary and doesn't seem to available via Suse.
 - Leaves Sbcl as last chance. According to their homepage, CLisp should
   actually work. Stop, I see they only mention "CLISP 2.33.2" ...
Anyway, I'll contact Sbcl.

> >
> > HOWEVER, anyway, the only point for us trying Sbcl is that we are
> > looking for an open-source Lisp (usable for Maxima) faster than Ecl.
> > And I got the impression that Sbcl can't be faster then Ecl, or
> > am I wrong?
> >
> 
> If the testsuite is any indication, sbcl and cmucl are faster than ecl.
> Can't remember if ccl is or not.  But also, it used to be that gcl ran the
> testsuite faster than any of these.  Don't know if that's true anymore.
> 
> However, I think the testsuite these days consists of a lot of number
> crunching code, so that might bias the results if you're doing symbolic
> stuff.
>

Thanks for this information! I'll try ... (and if I manage to install Sbcl,
I'll report on the speed w.r.t. (our) Maxima-applications).

Oliver