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

Re: A92: MultiTask System


>
> I also made some weeks ago a multitask kernel. It use normal fargo programs,
and works perfectly.
>
> So If you want to see my work I send you in this mail all the files of this
project. EVERYTHING is in assembly language, and there are all the source codes
>

   Yes, your kernel is exactly what is at the core of my kernel.  I've had to
solve all the problems you mentioned also, otherwise it wouldn't pass for my
project.
   - So far I've solved memory sharing between TI and the Kernel (for instance,
your kernel would not be able to create new variabes or delete old ones, since
that may move the stacks and programs.  This would cause a crash.
   - I've also solved the screen problem.  I'm working on a way to share an
'overlay' screen between the processes.  This screen lays on top of the
standard LCD screen.  Fargo programs will still be able to switch the LCD and
control it entirely, then when returning, the user may turn the overlay back
on.
   - I've solved the keyboard delema also.  By tail-patching the keyboard
interrupt, I can be sure to give keystrokes to the correct process.  A
semaphore controls access to this functionality.  (not written yet, but all
planned out)
   - I've also designed a method to load and unload programs into Xinu space
from TI space, making Xinu 100% compatable with Fargo.
   - Also, I'm designing a library system for processes which do not need the
assembly macros Fargo does, but adds a little overhead. (necissary since Fargo
libraries are allowed to move, but Xinu processes must always find their
libraries in the same place).  When unloaded, Xinu Libraries may move around,
but after loading, they are locked in Xinu space.

  And lots more.  I think you'll like it, though I don't think it'll ever
compete with the MacOS or Windows.. :-(

  On the down side, it is big and sucks up some computation time.  ~7K for the
kernel right now.  The demo itself uses ~6K (extra screen buffer and mask
buffer take up most of that).

  Of course, right after the semester ends, the TI-92+ and TI-89 will come out
making everything I just wrote incompatable.  But I'll try to write it as
upgradeable as possible.  Hopefully I'll be able to translate it to gcc for
wider distribution, but the Code Warrior platform is far superior for
development.

>
> And I want to wish GOOD LUCK to Bryan Christopher Turner, because I am sure
his program will work, and will be very good.
> So go on, and I am waiting for the first released version !!
> LONG LIFE TO Xinu !
>

Thanks for the encouragement!  It's all for fun really (the grade is just a
bonus!).

--Bryan
bcturner@eos.ncsu.edu
References: