A89: For people with non-TI m68k experience?? (linuxTI project)

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

A89: For people with non-TI m68k experience?? (linuxTI project)


 > I am trying to find out if there is anyone on this mailing list who has
 > experience with the m68k from other projects, and not just related to
 > the stupid calculator.

Well, I've been making a living for quite a few years designing
systems (HW&SW) containing all sorts of Motorola CPUs, including 
m68k variants. Would that qualify ? ;-)

 > F*ck the fact that it was built as a calculator.  It is a m68k based
 > computer.  I could list a dozen other similarly equipped systems that
 > were on the desktop.

This is true, but not too many ran unix, did it ? 
 
 > THink about it, when we were putting together our simple 68k based
 > systems a few (well more then that now) years ago, we would have given
 > an arm and a leg for this much RAM, or a prebuilt display and keyboard
 > with a BIOS to use it.  My first system, and my first z80 machine had
 > toggle switches and blinky lights for I/O!

Yup, this is true. But on a Z80 system you ran your own Word Best Monitor 
And Loader or, if you were very advanced, CP/M which was not much 
more than a primitive library of low-level I/O routines. The resources 
in this calc are quite remarkable, however, Linux's requirements are
fairly high compared to them. I can imagine that you can make a kernel 
and hack all tools to actually generate code that you can load and run 
without an MMU (although here goes your system stability) but what
would you exactly do with it ? Forget about compiling, assembling, 
linking on the calc itself - these tools have never been created with 
tight resources in mind. So, you will have a shell. Which one ? Bash ? 
Which is 300K *without* the dynamic libs it links ? And this is just 
the executable. Without MMU you can not really count on VM; even if
you could, you have no suitable swap device. Not to mention, that even 
if you had swap device and MMU, you couldn't still make on-demand paging
for the 68000 can't rerun an insn which caused a bus error.

So, you share that quarter of a meg of RAM between your disk and
actual RAM for execution of everything. What will you execute ? 
Not much, I guess.

Porting Linux onto this calc will waste an awful lot of resources and
will not give you what is great in Linux (or unix in general): the
stability, the virtual memory, the multi-user, protection schemes,
security, networking - basically nothing.

It would actually be much more fruitful IMHO to reverse engineer the 
HW as much as possible and write a system based on some free RT kernel
(which there are a few) and put a unix-ish API on top of it. 
Significantly less resources would produce a much more usable
programming environment. Weeding all the unnecessary baggage from
Linux and massage everything to fit into the calculator would require
more effort, I believe.

Regards,

Zoltan
References: