home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!think.com!sdd.hp.com!ux1.cso.uiuc.edu!roundup.crhc.uiuc.edu!boston.crhc.uiuc.edu!not-for-mail
- From: jchandy@crhc.uiuc.edu (John Chandy)
- Newsgroups: comp.sys.mac.hardware
- Subject: Re: 68000 & Lisa VM (was Re: Motorola's 68060 for the Mac)
- Date: 24 Jan 1993 15:56:23 -0600
- Organization: University of Illinois, Center for Reliable and High-Performance Computing
- Lines: 39
- Distribution: world
- Message-ID: <1jv3a7INN49t@boston.crhc.uiuc.edu>
- References: <D2150056.oaq0me@erics.infoserv.com> <77386@apple.apple.COM>
- NNTP-Posting-Host: boston.crhc.uiuc.edu
-
- In article <77386@apple.apple.COM> north@Apple.COM (Don North) writes:
- >In article <D2150056.oaq0me@erics.infoserv.com> erics@infoserv.com writes:
- >>I don't know how this was accomplished. When the 68000 first came out
- >>it was not designed to support virtual memory. At the company I was
- >>then working for, we did a lot of looking to see how we might get around
- >>the limitations that prevented implementing VM. Finally we decided that
- >>there were some instructions (notably the block move instructions I
- >>think) for which not enough information was saved on page fault to enable
- >>them to be restarted. Possibly Apple's compiler never generated these
- >>instructions? On the systems I was working on, we couldn't guarantee
- >>that users wouldn't program in assembly language and use these instructions.
-
- >The Lisa OS and H/W implemented a segmentation-based VM environment. It
- >was not a demand-paged system as are most today (ie, Mac VM in Apple's case).
- >References that might cause a fault - by convention ONLY intersegment
- >procedure calls - had to use a calling sequence which was KNOWN to be
- >restartable on a vanilla 68000. Demand paging of data references WAS NOT
- >supported (ie, would result in a bus error trap to the application). All
- >this, and a full multitasking OS and GUI on a 5 MHz 68000 in 1 Megabyte!
-
-
- Of course, there were UNIX workstations from Apollo that also
- supported demand paging on a vanilla 68000. I believe Apollo
- addressed the problem by using 2 68000's. The second one was only
- used to load the faulting page into memory and then the other
- processor could continue. I think the main processor was simply put
- into a wait state until the page was loaded. Thus no faults were
- incurred on the main processor.
-
- I don't know if HP or Sun used 68000's in their first workstations. I
- would guess not, since I think they both started with the 020.
-
- John Chandy
- j-chandy@uiuc.edu
-
-
-
-
-
-