Re: [Sbcl-devel] Re: Porting to another architecture (XScale)

[Christophe R. <cs...@ca...>]

Jan Rychter <ja...@ry...> writes:

>>>>>> "Jan" == Jan Rychter <ja...@ry...>:
>  Jan> A general question: how difficult would it be to make SBCL work on
>  Jan> ARM processors, in particular the Intel XScale PXA255 chips used
>  Jan> in the extremely cool Linux-based Sharp Zaurus palmtops
>  Jan> (http://www.ezaurus.com/)?
>
>  Jan> I'm looking for a general estimate, is it a:
>  Jan> a) I can do this before breakfast,
>  Jan> b) it will take a couple of days,
>  Jan> c) weeks of work and tough stuff,
>  Jan> d) hairy, scary, and there be dragons ahead.
>
>  Jan> Which is it?
>
> Hmm. Given the overwhelming response that I received in response to this
> question (the response consisted of a round number of messages), I'll
> conclude for the moment that the answer is d) and make a mental note to
> the extreme nastiness of the dragons.

FWIW, I didn't see the original message either. 

My answer depends slightly on your experience with the hardware.  It's
generally quite difficult to get sbcl running on a new platform (it's
easier if the cmucl people have done some of the hard work in
prehistory :-) but that difficulty goes up quite substantially if you
are unfamiliar with the platform in question (and exponentially if you
are unfamiliar with any platform at all).

It's never going to be (a) or (b), I fear, unless you are already
experienced both with your platform and with the inner workings of the
compiler; an understanding of how lisp code gets transformed into
machine code, and why that particular machine code, is fairly
essential, if only to understand why a particular sequence of
operations says "error too early in !COLD-INIT, halting."  I'd say
(c) in terms of overall effort, but you will probably find that you'll
need to take a week away from time to time to let ideas settle and let
your subconscious work on it.

On the other hand, these things get easier in collaboration; have a
read of the sbcl-internals cliki, the cmucl design document, and the
like.  Targeted questions will certainly be answered either here or on
#lisp IRC; and I seem to recall that someone else recently (Lars
Brinkhoff?) was experimenting with the ARM architecture.

I hope that helps somewhat,

Cheers,

Christophe
-- 
http://www-jcsu.jesus.cam.ac.uk/~csr21/       +44 1223 510 299/+44 7729 383 757
(set-pprint-dispatch 'number (lambda (s o) (declare (special b)) (format s b)))
(defvar b "~&Just another Lisp hacker~%")    (pprint #36rJesusCollegeCambridge)