home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.m68k
- Path: netcom.com!ludis
- From: ludis@netcom.com (Ludis Langens)
- Subject: Re: m68k Unix
- Message-ID: <ludisDLD4IK.JvL@netcom.com>
- Organization: It's here somewhere
- References: <BPBFXjs.nimitz@delphi.com> <4djhe8$gq1@vega.cs.pdx.edu> <4dkenp$df8@mailhub.scitec.com.au>
- Date: Thu, 18 Jan 1996 05:56:44 GMT
- Sender: ludis@netcom7.netcom.com
-
- In article <4dkenp$df8@mailhub.scitec.com.au> johns@rd.scitec.com.au (John Saunders) writes:
- >From my rusty old memory, the '451 wasn't a _paged_ MMU. It could translate
- >addresses around, I can remember things like logical address mask. I don't
- >think it had the ability to page fault on "page not present" so couldn't
- >support memory over commitment.
-
- According to the ancient data sheet I quickly skimmed, the 68451 can support
- virtual memory. The biggest limitiation is that the 68451 cannot use
- memory descriptions that are stored in external RAM and paged into the MMU
- on demand. In other words, all memory used by the currently running process
- must be mapped by the on-chip descriptors in the '451. (Presumably it is
- possible to page descriptors in and out using a software BERR handler -
- but that is not the intended way of using the '451.) To go beyond the
- limit of 32 descriptors in a '451, it is possible to combine multiple MMUs
- in parallel. Another result of the limitation is that task switching can
- get very slow because all the MMUs might need to be software updated during
- a switch.
-
- I seem to remember hearing that an old Altos used a 68010 and four 68451s
- to run System 3 Unix.
-
- --
- unsigned long Obfuscate(unsigned long i) {unsigned long t;
- Ludis Langens return i ? (t = Obfuscate(i >> 1), (t << 1) + (i & 1) +
- ludis@netcom.com (t + 858993459 >> 2 & 572662306) * 3) : 0;}
-