LispOS on a 68000

[Dwight Hughes]

David Tillman wrote:
> 
>     How feasible would a Lisp/Scheme OS be on a 68000?
> 
>     Please realize that I am in question-asking-mode here. I have no
>     idea how Lisp internals are implemented. I am thinking about putting
>     together a homebrew 68000 or 68030 and it would be nice to run
>     something besides Forth on it. My concerns are:
> 
>         a. Speed - If I wirewrap it, I will be stuck to around 8-10MHz.
>            Not exactly blazing. If I use a 68030, I will probably knock
>            out a couple PCBs for it and speed won't be quite so much an
>            issue.

I'd go with the 68030 at least -- and give a good look at using a 68040
instead.

>         b. Implementation - I don't know enough about Lisp to actually make
>            my own (not a pretty one, anyway). I could probably hack together
>            some Lisp/Scheme looking thing, but that wouldn't be quite the same.

Check out T - it is a Lispy Scheme or a Schemey Lisp (depending on who
you ask)
that compiles to native code and has a back end for a 680x0 available
IIRC. You can
get source code, doc, binaries, ... at:

http://128.2.242.152/afs/cs.cmu.edu/project/ai-repository/ai/lang/scheme/impl/t/

The good news and the bad news are the same: it is fully bootstrapped to
be written
completely in itself. To generate a new binary you need to have a
running T
system -- or be able to fake it in some fashion (perhaps write an simple
interpreter
for T to run on -- speed would not matter since you would throw it out
anyway once
you got T to compile itself).

T is not a toy however - the LMI K-machine OS was written in T (not for
a 680x0
though). See:

http://www.eval-apply.com/LispOS/kmachine.htm

>     Thoughts? Has this road been covered before? I would probably just
>     communicate with it by way of serial at first; a nice bitmapped display
>     wouldn't be too hard though.

Sounds like fun -- go for it.

-- Dwight