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

[MiNT] Thoughs about future (VM)

Finally I can post to the list again :-) (sorry for the mess again)

Few days ago we had discussion about FreeMiNT future. Let me add my 

I've been thinking about VM very much lately. The very importynt 
thing that keeps us back from Unix world and porting *any* (well, 
almost, but more than today) is lack of shared libraries (slb's does 
not count), and that is connected strongly with VM (no VM - very hard 
to implement shared libraries in a good way).

This is my proposition:

1. Leave ST-RAM not virtualized, less problems.
2. Keep all todays programms in ONE virtual machine, ie: FreeMiNT 
starts, creates virtual machine from what is in memory. Aditionaly 
every "old style" program runs in that virtual machine
3. Every "new style" program gets his own virtual machine
4. We keep a "window" between all virtual machines at a fixed addres 
(the same in all virtual machines). This will allow "new style" to 
use AES, and AES based protocols (even with pointers if care is 
taken). This would probable make new malloc() mode (for 
accesing this memory) usefull. This memory should lie in fast-ram 
(for speed).

How to recognize "old-style" programs from "new-style" programs? 
easy: by a flag in a program header.

Open: acces for "news style" programs to ST-RAM. I think it should be 
prohibited and we should give a /dev/* interface for accesing 
hardware instead.

1. Old software runs as it is running now
2. No need to write disk driver yet
3. New software should (?) be compatibile with old-style OS (ie. it 
should run under SingleTOS and MagiC if no specific FreeMiNt 
functions are needed.

1. Slowdown (like in any VM)
2. Some libraries (esp. AES libraries) should be rewritten or 
modified for "new-style" software.


Semper Fidelis

Adam Klobukowski