home *** CD-ROM | disk | FTP | other *** search
- Path: fwi.uva.nl!not-for-mail
- From: roode@fwi.uva.nl (Erik van Roode)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: Virtual memory on an 68000? Is it possible?
- Date: 18 Feb 1996 20:16:14 +0100
- Organization: FWI, University of Amsterdam
- Distribution: world
- Message-ID: <4g7ttu$en0@carol.fwi.uva.nl>
- References: <4et07r$lj9@wn1.sci.kun.nl> <42116497@phoenix.owl.de> <31222D21.5783@info.unicaen.fr> <4g2e90$267@oak73.doc.ic.ac.uk>
- NNTP-Posting-Host: carol.fwi.uva.nl
-
- mdf@doc.ic.ac.uk (Martin Frost) writes:
-
-
- >Frank Wille wrote:
- >> You will need a 68010 at a minimum to support virtual memory. The
- >> 68000 doesn't support instruction continuation and the information
- >> in the exception stack frame is not sufficient for simulating an
- >> access.
-
- >It would certainly be possible to build a hardware device to generate an NMI
- >when a certain range of addresses was accessed, and then the CPU could turn
- >those into disk reads via a subtask. The difficulty would be temporarily
- >stopping the task performing the read while you made the disk access. Oh,
- >and you still couldn't have more than 16Mb of address space (ie about 12 Mb
- >of total RAM, and even that much would be messy)
-
- Such a device is called Memory Management Unit. The main problem with an 68000
- and virtual memory is the fact that the program counter is increased before
- the actual instruction is executed. 68010 and higher versions save an old
- copy of the program counter so the faulting instructing can be restarted
- after handling the page fault. The lack of this saved program counter in
- 68000 makes it simpply impossible to implement hardware supported virtual
- memory.
-
- Erik
-
- --
- Erik van Roode Q : Did you ever use MSDOS ?
- University of Amsterdam A : Yes, but I did not inhale !
- Dept. of Computer Science
- Email: roode@fwi.uva.nl
-