kurt@tc.fluke.COM (Kurt Guntheroth) (11/11/87)
(1) Protected virtual memory. I will not buy ANY computer without this feature from now on. This is the critical feature that keeps a multitasking machine from crashing when a program goes crazy. With VM, I can reliably keep the system up for days. (2) SCSI port. Hard disks and backer-uppers are important. (3) 2 Megs. At least. Standard. 8 would be more like it. (4) Continuing software (OS & tools) enhancements. (5) Lowest possible price. I don't need a 68030, or even a 68020. I need a 68010 so I can have the VM I require. A 68020 WOULD be nice though. I don't waht a "workstation". I want a "home workstation". Amiga cannot compete against SUN, IBM, DEC, Apollo, and Apple for the engineering workstation market. They are too late. They should stay with the advanced home, video production, and arcade game market. That is where their competitive advantages lie. Other features I don't need. (1) 1024x1024 graphics. I can't justify $1K for a fancy color tube I can't use for anything else. I could go for a faster blitter and copper, which you need for this type of resolution, but I don't waht to trade away any existing feature. (2) 3 Million colors. I would like more colors, but not if I have to sacrifice any other video spec. Or software compatibility. (3) Hardware IBM emulation. I think the Amiga is living proof that this is not needed any longer to get good software. If the 3000 had a 68020 or better, a software emulator would be fast enough for WP and simple number crunching.
mwm@eris.BERKELEY.EDU (Mike (My watch has windows) Meyer) (11/13/87)
In article <2224@sputnik.COM> kurt@tc.fluke.COM (Kurt Guntheroth) writes:
<(1) Protected virtual memory. I will not buy ANY computer without this
< feature from now on. This is the critical feature that keeps a
< multitasking machine from crashing when a program goes crazy. With VM,
< I can reliably keep the system up for days.
Carefull. That's two features, one of them ambiguous. Virtual memory
means either "demand paging" (i.e. - the memory looks like it's there,
but may not be) or "mapped memory" (i.e. - the memory isn't where you
think it is, but it is in memory), depending on where you studied.
Protected memory means that you can't read and/or write and/or execute
the memory in question. It can be (and has been) done without virtual
memory of either flavor. However, you still have to have enough of an
MMU that mapping is nearly free. Since it buys you lots of other nice
things, it's usually worth doing (even Seymour Cray thinks so).
On the other hand, demand paging costs more than mapping, and buys you
soft failure when processes get larger than real memory.
Debate/questions should go to comp.arch.
<I don't need a 68030, or even a 68020. I need a 68010 so I can have
<the VM I require.
It sounded like you really wanted protected memory. As explained, this
doesn't require VM. The 68000 will support memory mapping just fine.
It has problems with demand paging - it doesn't even *have* a page
fault, much less do it right :-). It does have a line to use when it
tries to address memory that's not there. The stack frame doesn't have
sufficient information to let the OS restart/continue (choose one,
depending on religion) the faulting instruction after paging things
in. These didn't stop the sufficiently dedicated/devious from doing
demand paging on the 68000 (or two of them, in some cases).
If you're going to do demand paging, real memory quits affecting
what/how much you can run, and just makes things go faster.
<mike
--
Teddies good friend has his two o'clock feast Mike Meyer
And he's making Teddies ex girl friend come mwm@berkeley.edu
They mistook Teddies good trust ucbvax!mwm
Just for proof that Teddy was dumb. mwm@ucbjade.BITNET
daveh@cbmvax.UUCP (Dave Haynie) (11/13/87)
in article <5887@jade.BERKELEY.EDU>, mwm@eris.BERKELEY.EDU (Mike (My watch has windows) Meyer) says: > Keywords: Commodore A3000 Do it RIGHT! Oh, no, here I go again.... > <I don't need a 68030, or even a 68020. I need a 68010 so I can have > <the VM I require. > The 68000 will support memory mapping just fine. > It has problems with demand paging - it doesn't even *have* a page > fault, much less do it right :-). It does have a line to use when it > tries to address memory that's not there. No, not really. It has a few exceptions that can be used for this, but none of them work for what you want. That's certainly why "68010" appears above instead of "68000". > The stack frame doesn't have > sufficient information to let the OS restart/continue (choose one, > depending on religion) the faulting instruction after paging things > in. These didn't stop the sufficiently dedicated/devious from doing > demand paging on the 68000 (or two of them, in some cases). Well, I've seen the Apollo scheme. They use two 68000s, the second one tells the first one what it was doing when the page fault occurred. It was a reasonable kludge back then, but that was before the '010 existed. Since the '010 is a pin-compatible replacement for the '000, there's absolutely no problem today. And you still need an MMU. > If you're going to do demand paging, real memory quits affecting > what/how much you can run, and just makes things go faster. Sure enough. Like, all my Amigas have more real memory than any of our Apollos, and appear to run faster, even though the Apollos have much higher clock speeds. A DN3000 would change this, certainly, though that has an '020 and does virtual memory right. And probably has more on-board RAM as well. > <mike -- Dave Haynie Commodore-Amiga Usenet: {ihnp4|caip|rutgers}!cbmvax!daveh "The B2000 Guy" PLINK : D-DAVE H BIX : hazy "Computers are what happen when you give up sleeping" - Iggy the Cat
kurt@tc.fluke.COM (Kurt Guntheroth) (11/24/87)
<<In article <2224@sputnik.COM> kurt@tc.fluke.COM (Kurt Guntheroth) writes: <<(1) Protected virtual memory. I will not buy ANY computer without this << feature from now on. This is the critical feature that keeps a << multitasking machine from crashing when a program goes crazy. With VM, << I can reliably keep the system up for days. <mwm@eris.BERKELEY.EDU (Mike (My watch has windows) Meyer) writes: <Carefull. That's two features, one of them ambiguous. Virtual memory <means either "demand paging" (i.e. - the memory looks like it's there, <but may not be) or "mapped memory" (i.e. - the memory isn't where you True, two features, but I did say PROTECTED, VIRTUAL memory. I require protected memory so one task doesn't crash another (or the kernel). I require virtual memory so I am not limited to the size of my physical memory or in the number or configuration of programs I may run, either of which may be surprising over a period of days. I am aware that all I need do is buy 888 Mbytes of memory and I can run concurrently all the software ever written for the Amiga. But I can't afford 888 Mbytes. If I have virtual memory, I can 'run' any program, paying only a speed penalty if there is not enough real memory. As for people saying virtual memory operating systems are too inefficient, I suggest you read comp.sys.atari or Pournelle's article in Byte and see what they are saying about multitasking. A VM system is as efficient as a non-VM system under the same circumstances. Give me 2 Mbytes of real memory, let me load 1 Mbyte of code into it, and run it concurrently or non-concurrently. Then show me where the VM is slowing me down. Oh sure, if the OS or MMU is brain-dead it may take too long to context switch, but that doesn't indict the method. A program may start up FASTER under a demand paged VM than a non-VM system? With VM, you can start as soon as the first lines of code are loaded.
farren@gethen.UUCP (Michael J. Farren) (11/24/87)
In article <2320@sputnik.COM> kurt@tc.fluke.COM (Kurt Guntheroth) writes: >As for people saying virtual memory operating systems are too inefficient, I >suggest you read comp.sys.atari or Pournelle's article in Byte and see what >they are saying about multitasking. Are you joking? The day Pournelle shows any understanding of any concept in computing more sophisticated than interrupts, I'll swoon out of sheer astonishment. Also, the makeup of comp.sys.atari (which I do read) is fairly similar to the makeup of this group. I don't see what they have to say which is any more accurate than what people here have to say. -- ---------------- Michael J. Farren "... if the church put in half the time on covetousness unisoft!gethen!farren that it does on lust, this would be a better world ..." gethen!farren@lll-winken.arpa Garrison Keillor, "Lake Wobegon Days"