🚀 go-pugleaf

In article <17157@cbmvax.commodore.com>, Randell Jesup writes:

> In article <1991Jan6.032929.1813@mintaka.lcs.mit.edu> rjc@wookumz.ai.mit.edu (Ray Cromwell) writes:
> >
> > Virtual Memory shouldn't be a problem either. The OS merely checks for
> >the presence of an MMU and enables virtual mem (for FAST mem. Its impossible
> >to protect Chip mem from the blitter. And I can't imagine a machine
> >with an MMU that doesn't have fast ram)
> 
> 	Not so simple, at least in implementation.  Some things (for various
> reasons) should NOT be paged out.

   Certainly things like message ports and other time-critical stuff would
need to stay in physical memory.

   There's an interesting point in that the critical items are usually of
rather small size, while the real fillers (DTP documents, graphics, text
etc) are usually non-critical.

   The point is, the OS could (on MMU machines) set up an area for virtual
memory, and applications could ask explicitly for memory that can swapped
(MEMF_VIRT?). Thus, the real fillers could have the advantage, and the
programs that don't use much ram could get along without VM.

   I suppose code hunks could be marked non-critical as well?

   This would bring us most of the advantages of VM without breaking
anyone, if I'm not mistaken.

> 	A few things could be protected.  Not much without breaking things,
> and would require a lot of code and MMU tables per-process.

   Obviously a more complex thing to do. I wouldn't expect this to happen.


                                        Have a nice day        -Henrik
______________________________________________________________________________
| Henrik Clausen, Graffiti Data | If the Doors of Perception where cleansed, |
| ...{pyramid|rutgers}!cbmvax!  | Man would see Reality as it is - Infinite. |
\______cbmehq!adspdk!hclausen___|_________________________________W. Blake___/