Re: [MiNT] virtual memory

[Frank Naumann <fnaumann@freemint.de>]

Hello!

> >Yep, it just takes longer before the fragmentation becomes a problem.
> >Apart from that, correct me if I'm wrong, but it seems to me that when
> >you resolve fragmentation this way (one huge address space), bits of
> >allocated memory will be spread across the 4GB space which means you'll
> >be swapping all the time even though you're not actually using as much
> >RAM as you physically have. Or am I imagining this wrong?
> 
> Actually, no it will not.  But if I go with Frank's idea, the point
> is moot anyway.

There is one very important point that we forgotten completly in the
discussion until now.

With one address space we can never implement fork() in an efficient way.
We must do the same as the moment - copy all segments on every context
switch for every forked child and parent process. Or move at least all MMU
table entries to the right places.

Due to the fork() inefficency especially shell scrips are executed so
slowly at the moment as the most time is spent in memory copy actions.


Tschuess
   ...Frank

--
ATARI FALCON 040 // MILAN 060
-----------------------------------------
http://www.cs.uni-magdeburg.de/~fnaumann/
e-Mail: fnaumann@freemint.de