Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!dispose.news.demon.net!demon!news.demon.co.uk!demon!hearsay.demon.co.uk!user From: slavins.at.hearsay.demon.co.uk@localhost (Simon Slavin) Newsgroups: alt.folklore.computers Subject: OS with no distinction between RAM and HD ? Date: Sat, 04 Apr 1998 18:13:59 +0100 Organization: First Sirian Bank Message-ID: NNTP-Posting-Host: hearsay.demon.co.uk X-NNTP-Posting-Host: hearsay.demon.co.uk [194.222.24.177] Lines: 32 Okay, you guys seem to know everything ... I'm preparing for some lectures and trying to justify the distinction between main memory and backing-store. The majority of OSs today include both VM and disk caching. This blurs the distinction between the two: one uses disk as if it's RAM, the other uses RAM as if it's a disk. It's obvious that while we hold in our mind a very clear distinction between stuff-that's-lost-when-you-turn-off and memory-used-to-store-things-between-sessions, that's no longer how our computers really work. We also now have a situation where a large proportion of micros work better when they're turned-off rarely. When so many micros either run relatively stable OSes which don't require frequent power-cycling or are portables with efficient power-saving modes, more and more people leave their computers turned on all the time except when it's necessary to modify the OS or fiddle with hardware. This means that losing RAM contents is no longer something that happens every night. Was there ever an OS that just dealt with disk as if it's part of main memory ? How did the user have to change the way they worked ? Simon. -- Simon Slavin -- Computer Contractor | DAMN WHO MESSED WITH MY CAPSLOCK http://www.hearsay.demon.co.uk | KEY that's better. -- Geoff Lane Check email address for UBE-guard. | My s/ware deletes unread >3 UBEs/day.| Junk email not welcome at this site. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!netnews.com!ais.net!uunet!in3.uu.net!bulb.garlic.com!not-for-mail From: Anne & Lynn Wheeler Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 04 Apr 1998 11:35:10 -0700 Organization: South Valley Internet Lines: 40 Message-ID: References: Reply-To: Anne & Lynn Wheeler NNTP-Posting-Host: lynn-18.garlic.com X-Newsreader: Gnus v5.6.2/Emacs 19.34 some of the early single-level-store OS tended to have poor trashing characteristics ... not only did all data look like was part of memory it was used that way ... adaptive pattern algorithms didn't exist and there were no hints like file open/close. cross-over when memory was cheaper/faster than disk probably came some time in the 70s. virtual memory prior to that was a way of cramming only the used part of the program into real memory w/o the developer having to incure all the trouble of doing overlays. after the cross-over ... disk caches were a method of using memory to compensate for the fact that CPUs&memory were improving much faster than disks (in at least one processor line ... relative system performance of disks declined by a factor of five over a fifteen year period between 68 & 83 ... i.e. processor/memory improved by a factor of 50*, disk improved by a only factor of 10* ... therefor disk relative system performance declined by factor of 5*). I started noticing it around '77/'78 time-frame when trying to upgrade dynamic adaptive algorithms ... and "scheduling to the bottleneck" ... probability that filesystem was primary sysem bottleneck had signficantly increased in the ten years that i had been working in the area. -- -- Anne & Lynn Wheeler | lynn@garlic.com, finger for pgp key http://www.garlic.com/~lynn/ ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newsgate.cistron.nl!het.net!newsfeed2.uk.ibm.net!sackheads.org!ibm.net!europa.clark.net!207.172.3.49!feed2.news.erols.com!erols!news1.best.com!nntp2.ba.best.com!thvv From: thvv@best.com (Tom Van Vleck) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sat, 04 Apr 1998 12:40:04 -0800 Organization: Multicians Lines: 14 Message-ID: References: NNTP-Posting-Host: thvv.vip.best.com X-Trace: 891722405 17255 (none) 206.86.0.12 X-Newsreader: MT-NewsWatcher 2.4.1 Simon Slavin wrote: > Was there ever an OS that just dealt with disk as if it's > part of main memory ? How did the user have to change > the way they worked ? See http://www.best.com/~thvv/multics.html or http://www.city.ac.uk/~sh392/multics/multics.html in the UK There's usually a qualitative change at some point as you increase the size and complexity of a program. You cross the boundary from "easy" to "hard," and making progress on hard programs is eight times as slow as on easy ones. What Multics did was to move the boundary significantly, making many more programs easy. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.direct.ca!news-peer-west.sprintlink.net!news.sprintlink.net!Sprint!newsfeed.concentric.net!winternet.com!skypoint.com!not-for-mail From: "George R. Gonzalez" Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sat, 4 Apr 1998 15:53:36 -0600 Organization: SkyPoint Communications, Inc. Lines: 17 Message-ID: <6g6a41$5kj$1@shadow.skypoint.net> References: NNTP-Posting-Host: dial056.skypoint.net Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Newsreader: Microsoft Outlook Express 4.71.1712.3 X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3 IIRC the IBM System/38 took this to extremes, with just ONE apparent storage level. Even from the assembly language level you just had addressable "files". Of course there was a lot of microcode working underneath to make this happen. Also IIRC they took this very advanced architecture and hobbled it by making RPG the only "high level" language available. ROTFL^99? ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news-fra1.dfn.de!news-kar1.dfn.de!news-was.dfn.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.direct.ca!newshub1.home.com!news.home.com!news.rdc1.pa.home.com!not-for-mail From: David Ness Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sat, 04 Apr 1998 18:37:47 -0500 Organization: @Home Network Lines: 29 Message-ID: <3526C44B.34CC4A79@Home.Com> References: NNTP-Posting-Host: cc545395-a.lwmrn1.pa.home.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.02 [en] (WinNT; I) Leonard Fischer wrote: > > > > Some early computers had only rotating magnetic media as > the main memory, I think this was usually drum instead of > disk, one that I know of is the one that Knuth dedicated > his Art of Programming series to, IBM 604? On these > machines a great deal of manual effort was put into having > the next instruction show up under the read head at the > right time. Later the assembler took care of interleaving > the instructions with some attempt at optimal placement for > performance. > Close, but not quite. The 604 was, I think, a `Calculating Punch'. The `rotating drum' machine was the IBM 650, supposedly one of the first machines that was produced in quantities greater than 1,000. The assembler you refer to was SOAP (the Symbolic Optimum Assembly Program) and it used what was called the `1 1/2 address' nature of the machine to do as you suggested. Instructions were 10 digits: 2 Opcode, 4 Operand and 4 `Next instruction' which meant is was the only machine I programmed where instructions were not sequential. To get back to the original question about `disk' as `main memory', it seems to me that the IBM 305 had some sort of scheme that was pretty much that, but I never programmed it so I can't remember the details. Perhaps someone else will recall... ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newsgate.cistron.nl!het.net!newsfeed2.uk.ibm.net!sackheads.org!ibm.net!europa.clark.net!204.59.152.222!news-peer.gip.net!news.gsl.net!gip.net!newsxfer3.itd.umich.edu!news1.best.com!nntp2.ba.best.com!not-for-mail From: David E. Fox Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sat, 4 Apr 1998 20:40:32 -0800 Lines: 68 Message-ID: <98040420544900.03913@belvdere> References: NNTP-Posting-Host: belvdere.vip.best.com Mime-Version: 1.0 X-Trace: 891752193 933 dfox 206.86.0.12 X-Newsreader: KRN http://ultra7.unl.edu.ar FollowupTo: alt.folklore.computers On Sat, 04 Apr 1998, Simon Slavin wrote: >Okay, you guys seem to know everything ... > >I'm preparing for some lectures and trying to justify the >distinction between main memory and backing-store. The >majority of OSs today include both VM and disk caching. >This blurs the distinction between the two: one uses disk >as if it's RAM, the other uses RAM as if it's a disk. OSes such as Linux use both techniques, but disk caching is not at all the same thing as virtual memory. VM is just a technique wherein you remap an address using various things like page tables, and all addresses presented to the system are virtual, whether they reside in real memory or not. Of course, if the page where the address resides is not on disk, then the OS gets the page in from disk and restarts the instruction. Disk caching is just the idea of using multiple buffers to store disk contents; frequently-referenced contents hopefully, as it will be faster to refer to a particular disk area multiple times if it is actually in RAM rather than on disk. Linux, for instance, has a common 'pool' of available RAM that is used for processes, disk caching, and VM caching (it does actually cache frequently-used areas of VM as well, if I understand this correctly), and will use the pool for whatever needs it. If a process needs some ram, it gets it; same for disk and VM. >We also now have a situation where a large proportion of >micros work better when they're turned-off rarely. When I think that's a hardware thing -- it's still a good idea to keep the system running, because frequent power-cycling is stressful on the hardware. This really isn't an OS issue, except for the fact that OSes like Linux are really meant to be running all the time, which is different from other OSes where one tends to turn the system on, do whatever is needed in the app, and then turn the system off when finished. My first book on Unix I got (can't remember the title) when I got exposed to Linux some years ago made the point rather strongly of not just turning the system off when you're done. >This means that losing RAM contents is no longer >something that happens every night. True, but not really relevant. RAM is just a fast cache for virtual memory. >Was there ever an OS that just dealt with disk as if it's >part of main memory ? How did the user have to change >the way they worked ? The one that closest comes to mind is OS/400 on the midframe IBM AS/400 series. I use it in my job, but only as a user, not an administrator, but IIRC this OS does really blur the distinction between primary and 'secondary' storage. For Linux, the difference might be if you just had one big large partition on your system, and it was swap. Everything then would just be an address to the system, no matter where it came from. >Simon Slavin -- Computer Contractor | DAMN WHO MESSED WITH MY CAPSLOCK > http://www.hearsay.demon.co.uk | KEY that's better. -- Geoff Lane > Check email address for UBE-guard. | >My s/ware deletes unread >3 UBEs/day.| Junk email not welcome at this site. -- ------------------------------------------------------------------------ David E. Fox Tax Thanks for letting me dfox@belvdere.vip.best.com the change magnetic patterns root@belvedere.sbay.org churches on your hard disk. ----------------------------------------------------------------------- ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newsgate.cistron.nl!het.net!newsfeed2.uk.ibm.net!sackheads.org!ibm.net!news-lond.gip.net!news-peer.gip.net!news.gsl.net!gip.net!news-peer.sprintlink.net!news-backup-east.sprintlink.net!news-in-east.sprintlink.net!news.sprintlink.net!Sprint!204.156.128.20!news1.best.com!nntp2.ba.best.com!not-for-mail From: David E. Fox Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sat, 4 Apr 1998 20:57:54 -0800 Lines: 23 Message-ID: <98040421010201.03913@belvdere> References: <3526C44B.34CC4A79@Home.Com> NNTP-Posting-Host: belvdere.vip.best.com Mime-Version: 1.0 X-Trace: 891752562 933 dfox 206.86.0.12 X-Newsreader: KRN http://ultra7.unl.edu.ar FollowupTo: alt.folklore.computers On Sat, 04 Apr 1998, David Ness wrote: >To get back to the original question about `disk' as `main memory', >it seems to me that the IBM 305 had some sort of scheme that was >pretty much that, but I never programmed it so I can't remember the >details. Perhaps someone else will recall... I don't recall exactly what machine(s) might have done this, but I do happen to have a rather old(ish) book on computers (published '57 or '58) that describes one rather peculiar rotating drum arrangement. It was on a system where the drum was the main memory, but there was some special area where stuff was read from one area of the drum and continually deposited on the other, in some sort of endless ring arrangement. -- ------------------------------------------------------------------------ David E. Fox Tax Thanks for letting me dfox@belvdere.vip.best.com the change magnetic patterns root@belvedere.sbay.org churches on your hard disk. ----------------------------------------------------------------------- ###### Newsgroups: alt.folklore.computers Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news.gsl.net!gip.net!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!ix.netcom.com!lfischer From: lfischer@netcom.com (Leonard Fischer) Subject: Re: OS with no distinction between RAM and HD ? Message-ID: Organization: Trax Softworks X-Newsreader: TIN [version 1.2 PL2] References: Date: Sat, 4 Apr 1998 22:44:11 GMT Lines: 71 Sender: lfischer@netcom17.netcom.com Simon Slavin (slavins.at.hearsay.demon.co.uk@localhost) wrote: : Okay, you guys seem to know everything ... Ha! ...snip... : Was there ever an OS that just dealt with disk as if it's : part of main memory ? How did the user have to change : the way they worked ? Some early computers had only rotating magnetic media as the main memory, I think this was usually drum instead of disk, one that I know of is the one that Knuth dedicated his Art of Programming series to, IBM 604? On these machines a great deal of manual effort was put into having the next instruction show up under the read head at the right time. Later the assembler took care of interleaving the instructions with some attempt at optimal placement for performance. The IBM System 38, now superceded by the AS/400 provides a single model of memory, through a relational database { no access to raw memory or disk. Or so I've read, I haven't actually worked on it. Although in some ways your question - Was there ever an OS that just dealt with disk as if it's part of main memory ? How did the user have to change the way they worked ? is very clear, in other ways I'm not sure what you mean. All Virtual Memory operating systems try to present an application program with the illusion of main memory which is actually being simulated by the backing store. I'm not aware of any computer system that tried to present a permanent illusion of a larger main memory to the user by using the disk as main memory { it seems to me that this would have to take place "underneath" the operating system, either via hardware or microcode, otherwise at SOME point, particularly at initialization and whenever a page fault occurred SOMETHING has to do work to maintain the illusion. I think it's true that the technical reasons for a distinction between "main" store and "backing" store (accepting the British terms for the moment to be consistent between the 2, simply because I've never heard the term "backing memory") have changed a great deal over the years as the underlying technology available to implement storage has changed. Main store has been volatile (vacuum tubes, Mercury delay line), then non- volatile (drum, disk, core) and is now volatile again. The relative speed between main and backing store have also changed dramatically over the years and even from one computer system to another in the same timeframe depending on what engineering decisions (cost/performance tradeoffs) were made. Meanwhile computer architecture hasn't changed nearly as much. Probably because it's hard to change how we think. For a very interesting discussion of similar points in regards to the distinction between memory and processors, see the first few pages of the book, "The Connection Machine" by Daniel Hillis. He makes the point that in the early days of computers the processor was made of very expensive stuff { vacuum tubes, and the main memory was much less expensive { mercury delay lines and the like. Nowadays the processor and the memory are both made of silicon transistors, yet design decisions that were made then to optimize the usage of the processor are still in place today { hence the Von Neumann bottleneck. Len ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsm.ibm.net!ibm.net!newsjunkie.ans.net!newsfeeds.ans.net!news1.rmi.net!not-for-mail From: andyr@rmi.net (Andy Rabagliati) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 5 Apr 1998 18:21:03 GMT Organization: wizzy Lines: 9 Message-ID: <6g8i2f$cga$1@news1.rmi.net> References: NNTP-Posting-Host: webmark.com Bcc: andyr@wizzy.com Mail-Copies-To: never X-URL: http://www.wizzy.com/andyr X-Face: ",aVj->sw%,@..^N%][.Wl5q%zy@iYVP }s~5e7KK#0QtdE:I-sTp6{lBs4OoCa4i5qF.j?VoL&sEEu8'P=}(\NZl8\Ay'&]er&+'4X1e-`X;Hp k!T|X2(j%vW7$#-Watfta;30SjuX{c%&f->C,>w)=j|p2b[~Uupqn]WX7fWuk{eV%j)Dj;z6l~i!v0 I6AW/{etQls#xbOS4@:T2f8)5~JX>3erK2gJxvrvctC]^^&-LlTj6Q'cp;7j>?qW$P(1qhZX[)E>#= oOR'xy2*)"1GPpmU0L=[J'fC`RSxBJ%H. X-Newsreader: trn 4.0-test61 (25 January 1997) Originator: andyr@rmi.net (Andy Rabagliati) According to Simon Slavin : > > Was there ever an OS that just dealt with disk as if it's > part of main memory ? You can read "The Story of Mel" at http://www.wizzy.com/andyr/Mel.html to get a detailed look at the RPC-4000 drum memory computer. Cheers, Andy! ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!rill.news.pipex.net!pipex!dispose.news.demon.net!demon!news.demon.co.uk!demon!josstix.demon.co.uk!not-for-mail From: azz@josstix.demon.co.uk Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sun, 05 Apr 1998 20:06:38 GMT Message-ID: <891806798.9398.0.nnrp-07.c1edda47@news.demon.co.uk> NNTP-Posting-Host: josstix.demon.co.uk X-NNTP-Posting-Host: josstix.demon.co.uk [193.237.218.71] Lines: 10 This sounds remarkably like the machine described in "The Story Of Mel, A Real Programmer" that can be found in the Jargon File. Has this architecture ever been revived? Are there any modern machines which have a jump-to address after each instruction? And would it have any performance benefit? Thanks, Adam Sampson, azz@josstix.demon.co.uk ###### Date: 05 Apr 98 20:18:14 -0800 From: "Charlie Gibbs" Subject: Re: OS with no distinction between RAM and HD ? References: Message-ID: <1852.399T745T12183798@sky.bus.com> Newsgroups: alt.folklore.computers Lines: 41 X-Newsreader: THOR 2.5a (Amiga;TCP/IP) NNTP-Posting-Host: news.skybus.com Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news-peer.sprintlink.net!news-backup-west.sprintlink.net!news-in-west.sprintlink.net!news.sprintlink.net!Sprint!204.244.4.2!news.westel.com!news.skybus.com!204.244.247.106 In article slavins.at.hearsay.demon.co.uk@localhost (Simon Slavin) writes: >I'm preparing for some lectures and trying to justify the >distinction between main memory and backing-store. The >majority of OSs today include both VM and disk caching. >This blurs the distinction between the two: one uses disk >as if it's RAM, the other uses RAM as if it's a disk. >It's obvious that while we hold in our mind a very clear >distinction between stuff-that's-lost-when-you-turn-off >and memory-used-to-store-things-between-sessions, that's >no longer how our computers really work. Be careful here. Memory is not by definition volatile, even though currently the most cost-effective memories are. However, the core and plated-wire memories of 25 to 30 years ago would hold their contents when you powered down. The difference between memory and disk is that the data in memory can be directly operated on by the CPU, while data on disk has to be read or written to or from memory before it can be operated on. This is more consistent with the terms "memory" and "backing store" that you started off with. VM and RAM disks blur the distinction, but if you go right down to the metal it's still there. The programming model is different, though. >We also now have a situation where a large proportion of >micros work better when they're turned-off rarely. This is nothing new. Around 1983 I was working in a shop with a small mainframe and a HUGE air conditioner. We were shutting down the computer every night, and it was pretty flaky. It turns out that the thermal cycles were making the VLSI chips walk out of their sockets. Once we started leaving the machine on 24 hours a day, the random crashes went away. -- cgibbs@sky.bus.com (Charlie Gibbs) Remove the first period after the "at" sign to reply. ###### Date: 05 Apr 98 20:19:50 -0800 From: "Charlie Gibbs" Subject: Re: OS with no distinction between RAM and HD ? References: Message-ID: <895.399T2519T12195840@sky.bus.com> Newsgroups: alt.folklore.computers Lines: 25 X-Newsreader: THOR 2.5a (Amiga;TCP/IP) NNTP-Posting-Host: news.skybus.com Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news-peer.sprintlink.net!news-backup-west.sprintlink.net!news-in-west.sprintlink.net!news.sprintlink.net!Sprint!204.244.4.2!news.westel.com!news.skybus.com!204.244.247.106 In article thvv@best.com (Tom Van Vleck) writes: >Simon Slavin wrote: > >> Was there ever an OS that just dealt with disk as if it's >> part of main memory ? How did the user have to change >> the way they worked ? > >See http://www.best.com/~thvv/multics.html >or http://www.city.ac.uk/~sh392/multics/multics.html in the UK > >There's usually a qualitative change at some point >as you increase the size and complexity of a program. >You cross the boundary from "easy" to "hard," >and making progress on hard programs is eight times >as slow as on easy ones. What Multics did was to move >the boundary significantly, making many more programs easy. And then Intel came along and moved the boundary way down (64K segments), and many programs became hard. -- cgibbs@sky.bus.com (Charlie Gibbs) Remove the first period after the "at" sign to reply. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newscore.univie.ac.at!europa.clark.net!4.1.16.34!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!news.mindspring.net!gatech!nntp-xfer.ncsu.edu!not-for-mail From: cddukes@cc04du.unity.ncsu.edu (Christopher D Dukes) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 5 Apr 1998 21:12:32 GMT Organization: The Alfred Packer Memorial Dining Hall Lines: 10 Message-ID: <6g8s40$14c$1@uni00nw.unity.ncsu.edu> References: <6g6a41$5kj$1@shadow.skypoint.net> NNTP-Posting-Host: cc04du.unity.ncsu.edu X-Html-LART:

12:00

 X-UCE-Policy-00: Unsolicited Commercial E-mail sent to this account will X-UCE-Policy-01: be billed a $1000 editing fee. The sending of UCE to X-UCE-Policy-02: this account will be considered acceptance of these X-UCE-Policy-03: terms. OS/400 on the AS/400 continued the tradition. Although now one has access to real 'C' compilers and there is enough of a POSIX subsystem that perl has been ported. -- The following must be destroyed. Microsoft, Lyons Partnership. Balkanize USENET! Vote from the rooftops!! The best thing in RTP is now SMOG! Sending unsolicited commercial massmail to this account may result in a network outage for your site. Have a nice day. "Securing a Windows NT system -- Wire Cutters or Thermite?" ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!baron.netcom.net.uk!netcom.net.uk!server3.netnews.ja.net!ulcc.ac.uk!not-for-mail From: Kevin Ashley Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Mon, 06 Apr 1998 11:16:52 +0100 Organization: Posted via ULCC Internet Services Lines: 21 Message-ID: <3528AB94.41C6@ulcc.ac.uk> References: NNTP-Posting-Host: silver.ulcc.ac.uk Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.01 (X11; I; IRIX 5.3 IP22) To: Simon Slavin Simon Slavin wrote: ... > > Was there ever an OS that just dealt with disk as if it's > part of main memory ? How did the user have to change > the way they worked ? > I used to be regaled by various folk who had worked at EUCS (Edinburgh University Computing Services) in the 1970s who described EMAS as working in this way. There was one huge address space which took in memory and the filestore - at least that's how they described it. Never having used EMAS myself, I can't be sure. ------------------------------------------------------------------------ Kevin Ashley K.Ashley@Ulcc.ac.uk Special Projects Manager http://www.ulcc.ac.uk/staff/Kevin+Ashley ULCC ...ukc!ncdlab!K.Ashley (but probably not any more) This is not a signature ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!howland.erols.net!frankfurt.de.uu.net!newsfeed.nacamar.de!uni-erlangen.de!cs.tu-berlin.de!not-for-mail From: Matthias Heidbrink <*NOSPAM*mh@cs.tu-berlin.de> Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Mon, 06 Apr 1998 22:11:49 +0100 Organization: Technical University of Berlin, Germany Lines: 41 Message-ID: <35294515.1234@cs.tu-berlin.de> References: Reply-To: *NOSPAM*mh@cs.tu-berlin.de NNTP-Posting-Host: mh.isdn.cs.tu-berlin.de Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.0 (OS/2; I) Hi, Simon Slavin wrote: > Was there ever an OS that just dealt with disk as if it's > part of main memory ? How did the user have to change > the way they worked ? I know of two systems that worked with Fixpoints, "Eumel" and later "Ergos L3". Both are multiuser multitasking systems developed by Jochen Liedke at GMD, a German Computer Science research instutute. I have never worked with it, but it must have been very interesing. "Fixpoints" means that the whole system state is written to disk every few minutes. As this mechanism was directly connected with the memory management system, it didn't cost much overhead. If the system ever crashed or was switched off without making a clean shutdown, it always had the state of the time when the last fixpoint was written when it came up again. "Eumel" was used at German Schools for the education in Computer Science. That must have been 15 Years ago, not my time. Eumel had memory protection even on computers like the Z80 by working with a MMU in software. Someone told me that a 4 Mhz Z80 system with 5 terminals connected had a better response time than the PCs of these days - on every terminal at the same time, of course ;-)) . "Ergos L3" was based on the L3 Microkernel for the 386. This microkernel was so efficient that device drivers could be implented as user processes. See "http://www.aci.technopark.gmd.de/ACI/aciprod/L3_H.html". There's even a free version of "Ergos L3" for playing around with it. You can find information about the L3 Microkernel under "http://os.inf.tu-dresden.de/L4/". Ciao, Matthias ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.internetmci.com!204.59.152.222!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!baron.netcom.net.uk!netcom.net.uk!server3.netnews.ja.net!news.ox.ac.uk!not-for-mail From: Ben Hutchings Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 6 Apr 1998 22:40:10 +0100 Organization: Not organised Lines: 25 Sender: womble@eldritch.dyn.ml.org Message-ID: <6gbi3q$3ql$1@eldritch.dyn.ml.org> References: <35294515.1234@cs.tu-berlin.de> NNTP-Posting-Host: max54.public.ox.ac.uk In article <35294515.1234@cs.tu-berlin.de>, Matthias Heidbrink <*NOSPAM*mh@cs.tu-berlin.de> wrote: >Eumel had memory protection even on computers like the Z80 by working >with a MMU in software. That's hardly memory protection - unless all user programs were interpreted. >Someone told me that a 4 Mhz Z80 system with 5 terminals connected >had a better response time than the PCs of these days - on every >terminal at the same time, of course ;-)) . > >"Ergos L3" was based on the L3 Microkernel for the 386. This microkernel >was so efficient that device drivers could be implented as user >processes. AmigaOS lets you do that too. In fact, virtually the entire OS runs in user state. The drawback is of course the complete lack of memory protection. -- Ben Hutchings | ICOA User Rep election 1998: http://www.jms.org/election/ email/finger: m95bwh@ecs.ox.ac.uk | web: http://users.ox.ac.uk/~worc0223/ Absolutum obsoletum. (If it works, it's out of date.) - Stafford Beer ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!dispose.news.demon.net!demon!peer.news.zetnet.net!zetnet.co.uk!not-for-mail From: amroth@zetnet.co.uk.NOJUNK (Phil Edwards) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Mon, 06 Apr 1998 23:34:08 GMT Message-ID: <352931e2.88429@news.zetnet.co.uk> References: NNTP-Posting-Host: man-168.dialup.zetnet.co.uk Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/16.451 Lines: 15 slavins.at.hearsay.demon.co.uk@localhost (Simon Slavin) wrote: >Was there ever an OS that just dealt with disk as if it's >part of main memory ? How did the user have to change >the way they worked ? Get hold of Frank Soltis' _Inside the AS/400_ and look up "single-level store". Tell you what, I'll do it myself when I next resume my mysterious daytime identity of AS/400 Man. Phil -- Phil Edwards http://www.users.zetnet.co.uk/amroth "When I say 'serious' I mean like children are at play" - Peter Blegvad ###### Date: 07 Apr 98 18:15:05 -0800 From: "Charlie Gibbs" Subject: Re: OS with no distinction between RAM and HD ? References: <895.399T2519T12195840@sky.bus.com> Message-ID: <779.401T292T10953149@sky.bus.com> Newsgroups: alt.folklore.computers Lines: 21 X-Newsreader: THOR 2.5a (Amiga;TCP/IP) NNTP-Posting-Host: news.skybus.com Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news-peer.sprintlink.net!news-backup-east.sprintlink.net!news-in-east.sprintlink.net!news.sprintlink.net!Sprint!204.244.4.2!news.westel.com!news.skybus.com!204.244.247.101 In article bmarcum@iglou.com (bmarcum) writes: >On 1998-04-05 cgibbs@sky.bus.com said: > > >And then Intel came along and moved the boundary way down > >(64K segments), and many programs became hard. > >The 8086 with its 64k segments was a step up from its predecessor >which only addressed 64k total memory. Not much of a step up from bank-switched 8080 and Z-80 systems, which could also address up to a megabyte of memory. As a programmer, I find the infamous 640K barrier to be much less an impediment than the 64K barrier which was still imposed by the inherent 16-bitness of the 8086's addressing. Intel put the "backward" in "backward-compatible". -- cgibbs@sky.bus.com (Charlie Gibbs) Remove the first period after the "at" sign to reply. ###### Newsgroups: alt.folklore.computers Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newsgate.cistron.nl!het.net!peer.news-uk.wisper.net!peer.news-uk.wisper.net!netnews.com!news.maxwell.syr.edu!newsfeed.internetmci.com!192.107.41.5!iglou!news From: bmarcum@iglou.com Subject: Re: OS with no distinction between RAM and HD ? X-Nntp-Posting-Host: lou-ts7-28.iglou.com X-Everything: Net-Tamer V 1.08X X-Nntp-Posting-User: [unauthenticated] Message-ID: Sender: news@iglou.com (News) Organization: IgLou Internet Services (1-800-436-4456) References: <895.399T2519T12195840@sky.bus.com> Date: Tue, 7 Apr 1998 23:28:09 GMT Lines: 8 On 1998-04-05 cgibbs@sky.bus.com said: >And then Intel came along and moved the boundary way down >(64K segments), and many programs became hard. The 8086 with its 64k segments was a step up from its predecessor which only addressed 64k total memory. Net-Tamer V 1.08X - Test Drive ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.nacamar.de!uni-erlangen.de!cs.tu-berlin.de!not-for-mail From: Matthias Heidbrink <*NOSPAM*mh@cs.tu-berlin.de> Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Wed, 08 Apr 1998 03:38:53 +0100 Organization: Technical University of Berlin, Germany Lines: 31 Message-ID: <352AE33D.38D4@cs.tu-berlin.de> References: <35294515.1234@cs.tu-berlin.de> <6gbi3q$3ql$1@eldritch.dyn.ml.org> Reply-To: *NOSPAM*mh@cs.tu-berlin.de NNTP-Posting-Host: mh.isdn.cs.tu-berlin.de Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.0 (OS/2; I) Hi, Ben Hutchings wrote: > >Eumel had memory protection even on computers like the Z80 by working > >with a MMU in software. > > That's hardly memory protection - unless all user programs were > interpreted. Don't ask me how it worked. Possible P-Code. The compiler on this system was Elan (looks similiar to Pascal and C). > >"Ergos L3" was based on the L3 Microkernel for the 386. This microkernel > >was so efficient that device drivers could be implented as user > >processes. > > AmigaOS lets you do that too. In fact, virtually the entire OS runs > in user state. > The drawback is of course the complete lack of memory > protection. The difference is that L3 does this _with_ full memory protection. The L3 and L4 microkernels have an IPC performance that is many times better than of systems like Mach, otherwise this would cost to much performance. Both are written in highly-optimized assembler. L3 is optimized for the Intel 386. L4 is portable, but what I read about it looks like that only the Kernel API and concepts are reused and the kernel itself is rewritten for every processor type. Ciao, Matthias ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news.gsl.net!gip.net!howland.erols.net!woodstock.news.demon.net!demon!news.demon.co.uk!demon!hearsay.demon.co.uk!user From: slavins.at.hearsay.demon.co.uk@localhost (Simon Slavin) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Thu, 09 Apr 1998 00:39:07 +0100 Organization: First Sirian Bank Message-ID: References: <35294515.1234@cs.tu-berlin.de> NNTP-Posting-Host: hearsay.demon.co.uk X-NNTP-Posting-Host: hearsay.demon.co.uk:194.222.24.177 X-Trace: news.demon.co.uk 892078746 nnrp-02:23923 NO-IDENT hearsay.demon.co.uk:194.222.24.177 X-Complaints-To: abuse@demon.net Lines: 34 In article <35294515.1234@cs.tu-berlin.de>, Matthias Heidbrink <*NOSPAM*mh@cs.tu-berlin.de> wrote: > Simon Slavin wrote: > > Was there ever an OS that just dealt with disk as if it's > > part of main memory ? How did the user have to change > > the way they worked ? > > I know of two systems that worked with Fixpoints, "Eumel" and later > "Ergos L3". Both are multiuser multitasking systems developed by Jochen > Liedke at GMD, a German Computer Science research instutute. > > I have never worked with it, but it must have been very interesing. > "Fixpoints" means that the whole system state is written to disk every > few minutes. As this mechanism was directly connected with the memory > management system, it didn't cost much overhead. > If the system ever crashed or was switched off without making a clean > shutdown, it always had the state of the time when the last fixpoint was > written when it came up again. Beautiful. That's exactly the kind of thing I was asking about. Thanks for posting it. I'm hoping to come across a model where 'files' may well exist in memory but not on disk and where 'working space' could be on disk instead of memory. Simon. -- Simon Slavin -- Computer Contractor | DAMN WHO MESSED WITH MY CAPSLOCK http://www.hearsay.demon.co.uk | KEY that's better. -- Geoff Lane Check email address for UBE-guard. | My s/ware deletes unread >3 UBEs/day.| Junk email not welcome at this site. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newscore.univie.ac.at!uninett.no!news.algonet.se!feed1.news.luth.se!luth.se!news.kth.se!escapen.nada.kth.se!d96-abo From: d96-abo@escapen.nada.kth.se (Alexander Boström) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 9 Apr 1998 09:04:56 GMT Lines: 21 Message-ID: <6gi2vo$gkm$1@news.kth.se> References: NNTP-Posting-Host: escapen.nada.kth.se Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Newsreader: TIN [version 1.2 PL2] Simon Slavin (slavins.at.hearsay.demon.co.uk@localhost) wrote: : Was there ever an OS that just dealt with disk as if it's : part of main memory ? How did the user have to change : the way they worked ? You should have a look at the OS:es that gets rid of files, and instead just considers everything an object (a C "struct", a LISP "pair", a Java "object"). The state of the system is somehow preserved through a power down-up cycle. A simple interpreter for some language plus some fixpoint thingy is one example. Hey, just look at SmallTalk, it has all the properties you describe! Another ex.: http://www.tunes.org/ -- Alexander http://www.nada.kth.se/~d96-abo/index-en.html d96-abo@nada.kth.se ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!feed1.news.luth.se!luth.se!masternews.telia.net!news-feed.inet.tele.dk!bofh.vszbr.cz!newspeer.monmouth.com!uunet!in4.uu.net!news.uiowa.edu!not-for-mail From: jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 9 Apr 1998 13:52:55 GMT Organization: The University of Iowa Lines: 38 Message-ID: <6gijrn$f7o$1@flood.weeg.uiowa.edu> References: NNTP-Posting-Host: pyrite.cs.uiowa.edu Simon Slavin wrote: > Was there ever an OS that just dealt with disk as if it's > part of main memory ? How did the user have to change > the way they worked ? Atlas, built by Feranti, in the early 1960's, did this. To the programmer, the disk, core memory and even a large part of the magnetic tape library all appeared as a single large RAM. Multics, built by a consortium of Bell Labs, MIT and GE (but Bell dropped out and GE sold its computer business to Honeywell) did this too. To the programmer, open files were just segments of memory. Thus, from a C and UNIX perspective, the open() and shmat() operations were identically the same! Multics was built in the late 1960's and commercially available throughout the 1970's and into the 1980's. Least anyone think that Multics was a dead end project, the legacy of Multics lives on. Multics was built in order to explore the idea of an information utility. Today, information utilities, to which the user connects by modem and pays a monthly service charge, are commonplace, and every one of them owes a debt to Multics! By the 1970's mainframe computer systems routinely used the virtual memory ideas that began with UNIX. Typically, a partition of the hard disk system was used to extend the system's RAM. The antique UNIX workstation from which I'm posting this works this way. It only has 8 meg of RAM, but I routinely run programs that need as much as 10 or 15 meg, and it doesn't complain, it just runs them a bit slower than it would if I had more RAM. Today, my understanding is that MAC OS and Windows 95 can be configured to do this too, although this is a recent development. Unlike Multics, though, these systems don't let you open files by mapping the file into the memory address space. The last system I know of that supported that beautiful idea was BSD 4.3 UNIX, and hardly anyone ever used that feature. Sadly, my antique UNIX system is running BSD 4.2, so I can't play with that. Doug Jones jones@cs.uiowa.edu ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news.gsl.net!gip.net!newshub1.home.com!news.home.com!peerfeed.ncal.verio.net!nntp.ni.net!bhahn From: bhahn@spam-spam.go-away.com (Brendan Hahn) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Thu, 09 Apr 1998 15:32:32 -0800 Organization: Transoft Corp Lines: 16 Message-ID: References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> <352ef8ff.1461279@news.innet.be> <6gj3lr$557$1@eldritch.dyn.ml.org> Reply-To: bhahn@transoft.mangle.net (unmangle address to reply) NNTP-Posting-Host: host6.transoft.net Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit X-Newsreader: Yet Another NewsWatcher 2.4.0 Ben Hutchings wrote: >In article <352ef8ff.1461279@news.innet.be>, >Luc Van der Veken wrote: >> >>I don't know about MACOS, but both windows 95 and NT know memory >>mapped files. > >As do the UNIX and UNIX-alike systems I've used, which are not based >on BSD. Linux actually loads executables by simply mapping them into >a new process's memory. Macs have it too, though I'm not sure if it's a service directly available to programmers. It may be only an internal element of the process management and shared library systems. bhahn@transoft.mangle.net <-- unmangle to reply ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!nntp.abs.net!Supernews60!supernews.com!Supernews73!supernews.com!Supernews69!not-for-mail From: die.spam@hell.org.us (Evandro Menezes) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Thu, 09 Apr 1998 16:14:48 GMT Organization: E&R Lines: 19 Message-ID: <352cf21a.7865085@news.nabi.net> References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> Reply-To: evandro@geocities.com (Evandro Menezes) NNTP-Posting-Host: 13936@208.6.184.164 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/32.451 In <6gijrn$f7o$1@flood.weeg.uiowa.edu>, jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) wrote: >Today, my understanding is that MAC OS and Windows 95 can be configured to >do this too, although this is a recent development. Unlike Multics, though, >these systems don't let you open files by mapping the file into the memory >address space. The last system I know of that supported that beautiful >idea was BSD 4.3 UNIX, and hardly anyone ever used that feature. Sadly, my >antique UNIX system is running BSD 4.2, so I can't play with that. Just for the records, although I am NOT an advocate of these systems, Windows 95 and NT CAN use memory mapped files. HTH ____________________________________________________________ Evandro Menezes Austin, TX USA Tel:+1-512-502-9199 ICQ:7957253 mailto:evandro@geocities.com http://over.to/evandro ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!nntp.abs.net!howland.erols.net!woodstock.news.demon.net!demon!news.demon.co.uk!demon!mail2news.demon.co.uk!tnglwood.demon.co.uk!unclebob From: Robert Billing Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Thu, 09 Apr 98 16:57:30 GMT Message-ID: <892141050snz@tnglwood.demon.co.uk> References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> Reply-To: unclebob@tnglwood.demon.co.uk X-Mail2News-User: unclebob@tnglwood.demon.co.uk X-Mail2News-Path: post-20.mail.demon.net!post.mail.demon.net!tnglwood.demon.co.uk X-Trace: mail2news.demon.co.uk 892143050 15816 unclebob tnglwood.demon.co.uk X-Complaints-To: abuse@demon.net X-Newsreader: Demon Internet Simple News v1.29 Lines: 19 In article <6gijrn$f7o$1@flood.weeg.uiowa.edu> jones@pyrite.cs.uiowa.edu "Douglas W. Jones,201H MLH,3193350740, " writes: > do this too, although this is a recent development. Unlike Multics, though, > these systems don't let you open files by mapping the file into the memory > address space. The last system I know of that supported that beautiful I have in fact done this on VMS. It is particularly useful with largeish files, such as uncompressed video clips. -- I am Robert Billing, Christian, inventor, traveller, cook and animal lover, I live near 0:46W 51:22N. http://www.tnglwood.demon.co.uk/ "Bother," said Pooh, "Eeyore, ready two photon torpedoes and lock phasers on the Heffalump, Piglet, meet me in transporter room three" ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed2.uk.ibm.net!sackheads.org!ibm.net!news-lond.gip.net!news.gsl.net!gip.net!baron.netcom.net.uk!netcom.net.uk!rill.news.pipex.net!pipex!bore.news.pipex.net!pipex!join.news.pipex.net!pipex!krypton.inbe.net!INbe.net!not-for-mail From: lucvdv@null.net (Luc Van der Veken) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Thu, 09 Apr 1998 17:22:44 GMT Organization: . Lines: 10 Message-ID: <352ef8ff.1461279@news.innet.be> References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> NNTP-Posting-Host: pool02b-194-7-226-45.uunet.be Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/32.451 X-No-Archive: yes jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) told us > ... MAC OS and Windows 95 ... Unlike Multics, though, > these systems don't let you open files by mapping the file into the memory > address space. I don't know about MACOS, but both windows 95 and NT know memory mapped files. ###### Newsgroups: alt.folklore.computers Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news-peer.sprintlink.net!news-peer-east.sprintlink.net!news.sprintlink.net!Sprint!ix.netcom.com!netcom!alderson From: alderson@netcom.netcom.com (Richard M. Alderson III) Subject: Re: OS with no distinction between RAM and HD ? In-Reply-To: jones@pyrite.cs.uiowa.edu's message of 9 Apr 1998 13:52:55 GMT Message-ID: Sender: alderson@netcom.netcom.com Reply-To: alderson@netcom.com Organization: NETCOM On-line services References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> Date: Thu, 9 Apr 1998 18:14:58 GMT Lines: 26 In article <6gijrn$f7o$1@flood.weeg.uiowa.edu> jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: >By the 1970's mainframe computer systems routinely used the virtual memory >ideas that began with UNIX. Virtual memory was already in use on PDP-10 systems in 1968, and was old tech- nology even then. It simply hadn't been discovered by IBM as yet... >Unlike Multics, though, these systems don't let you open files by mapping the >file into the memory address space. The last system I know of that supported >that beautiful idea was BSD 4.3 UNIX, and hardly anyone ever used that >feature. Tops-20 has had memory-mapping of files since 1976, I think as a legacy of TENEX (1972), as an integrated part of the OS: String- and byte-oriented I/O to/from disk is done by mapping multiple pages of the file and manipulating those. I'd always thought that the 4.3bsd memory-mapped files were a kludge based on a clean design from the 36-bit world. -- Rich Alderson You know the sort of thing that you can find in any dictionary of a strange language, and which so excites the amateur philo- logists, itching to derive one tongue from another that they know better: a word that is nearly the same in form and meaning as the corresponding word in English, or Latin, or Hebrew, or what not. --J. R. R. Tolkien, alderson@netcom.com _The Notion Club Papers_ ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-stock.gip.net!news-peer.gip.net!news.gsl.net!gip.net!newsfeed.nacamar.de!uni-erlangen.de!cs.tu-berlin.de!not-for-mail From: Matthias Heidbrink <*NOSPAM*mh@cs.tu-berlin.de> Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Thu, 09 Apr 1998 19:01:04 +0100 Organization: Technical University of Berlin, Germany Lines: 10 Message-ID: <352D0CE0.18C5@cs.tu-berlin.de> References: <35294515.1234@cs.tu-berlin.de> Reply-To: *NOSPAM*mh@cs.tu-berlin.de NNTP-Posting-Host: mh.isdn.cs.tu-berlin.de Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.0 (OS/2; I) Hi Simon, > I'm hoping to come across a model where 'files' may well exist > in memory but not on disk and where 'working space' could be on > disk instead of memory. Ergos L3 also supports Memory-mapped files. But I don't know exactly if all files are memory-mapped or also "ordinary" files exist. Ciao, Matthias ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!feed1.news.luth.se!luth.se!fci-se!fci!masternews.telia.net!news-stkh.gip.net!news-lond.gip.net!news.gsl.net!gip.net!baron.netcom.net.uk!netcom.net.uk!server3.netnews.ja.net!news.ox.ac.uk!not-for-mail From: Ben Hutchings Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 9 Apr 1998 19:22:51 +0100 Organization: Not organised Lines: 21 Sender: womble@eldritch.dyn.ml.org Message-ID: <6gj3lr$557$1@eldritch.dyn.ml.org> References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> <352ef8ff.1461279@news.innet.be> NNTP-Posting-Host: max80.public.ox.ac.uk In article <352ef8ff.1461279@news.innet.be>, Luc Van der Veken wrote: >jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H >MLH,3193350740,3193382879) told us > >> ... MAC OS and Windows 95 ... Unlike Multics, though, >> these systems don't let you open files by mapping the file into the memory >> address space. > >I don't know about MACOS, but both windows 95 and NT know memory >mapped files. As do the UNIX and UNIX-alike systems I've used, which are not based on BSD. Linux actually loads executables by simply mapping them into a new process's memory. -- Ben Hutchings | ICOA User Rep election 1998: http://www.jms.org/election/ email/finger: m95bwh@ecs.ox.ac.uk | web: http://users.ox.ac.uk/~worc0223/ Lowery's Law: If it jams, force it. If it breaks, it needed replacing anyway. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!nntp.abs.net!europa.clark.net!199.0.154.56!ais.net!uunet!in5.uu.net!news.uiowa.edu!not-for-mail From: jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 9 Apr 1998 21:17:16 GMT Organization: The University of Iowa Lines: 11 Message-ID: <6gjdss$o7m$1@flood.weeg.uiowa.edu> References: <6gj3lr$557$1@eldritch.dyn.ml.org> NNTP-Posting-Host: pyrite.cs.uiowa.edu From article <6gj3lr$557$1@eldritch.dyn.ml.org>, by Ben Hutchings : > As do the UNIX and UNIX-alike systems I've used, which are not based > on BSD. Linux actually loads executables by simply mapping them into > a new process's memory. I find no evidence of this in the pure System V derivitives. I agree that it's the right way to go, and I'm glad to hear that Linux does it! Doug Jones jones@cs.uiowa.edu ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news.gsl.net!gip.net!uwm.edu!chicago-news-feed2.bbnplanet.com!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!news.mindspring.net!cssun.mathcs.emory.edu!hubcap.clemson.edu!sc.edu!snoopy.uscsumter.edu!trev From: trev@sc.edu (Trevor Zion Bauknight) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Thu, 09 Apr 1998 22:36:07 -0400 Organization: Metaphysical Bunko Squad Lines: 28 Message-ID: References: <6gi2vo$gkm$1@news.kth.se> NNTP-Posting-Host: nas-sa-p4.usc.net X-Newsreader: MT-NewsWatcher 2.4.1 In article <6gi2vo$gkm$1@news.kth.se>, d96-abo@escapen.nada.kth.se (Alexander Boström) wrote: > You should have a look at the OS:es that gets rid of files, and instead > just considers everything an object (a C "struct", a LISP "pair", a Java This makes me think about those dedicated FORTH systems a few people used to sell, with the bitslice processors. FORTH is a language, but it had a lot of OS capabilities when it got to a certain level of development. I remember seeing these when I used to get FORTH Dimensions... I always wanted one of those really utilitarian boxes with the handle on top (looked like an army radio box or something)...think it was made by Hartronix or some such. Now, I guess, FORTH is primarily used in imbedded situations and in low-level stuff like Open Firmware because of the ease of porting it to new architectures... Is this what you mean? I'm no expert, but programming-wise, FORTH is the only thing (other than Z80 assembly, reluctantly) that I ever spent any time at all with... Trev -- http://www.uscsu.sc.edu/~tzbaukni trev@sc.eduEATMOSPAM (unmunge) ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newsgate.cistron.nl!het.net!newsfeed2.uk.ibm.net!sackheads.org!ibm.net!news-lond.gip.net!news-peer.gip.net!news.gsl.net!gip.net!howland.erols.net!agate!islay.ssl.berkeley.edu!korpela From: korpela@islay.ssl.berkeley.edu (Eric J. Korpela) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 9 Apr 1998 23:15:37 GMT Organization: Cal Berkeley-- Space Sciences Lab Lines: 25 Message-ID: <6gjkqp$g8u$1@agate.berkeley.edu> References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> NNTP-Posting-Host: islay.ssl.berkeley.edu In article <6gijrn$f7o$1@flood.weeg.uiowa.edu>, Douglas W. Jones,201H MLH,3193350740,3193382879 wrote: >Today, my understanding is that MAC OS and Windows 95 can be configured to >do this too, although this is a recent development. Unlike Multics, though, >these systems don't let you open files by mapping the file into the memory >address space. The last system I know of that supported that beautiful >idea was BSD 4.3 UNIX, and hardly anyone ever used that feature. On the contrary, most modern Unices support memory mapped files. An OS without memory mapped files is considered by many to be broken. Memory mapped files were possible in OS/2 version 2.X by hooking into the page fault handler. I was pissed off when that ability was removed from OS/2 version 3.X. Hell, even Win32 has memory mapped files. For more info, see the mmap() API in the Single Unix Specification: http://www.rdg.opengroup.org/onlinepubs/7908799/xsh/mmap.html The win32 API is CreateFileMapping() http://premium.microsoft.com/msdn/library/sdkdoc/pdwbase/fmap3_79wn.htm. Eric -- Eric Korpela | An object at rest can never be korpela@ssl.berkeley.edu | stopped. Click for home page. ###### Newsgroups: alt.folklore.computers Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!ix.netcom.com!lfischer From: lfischer@netcom.com (Leonard Fischer) Subject: Re: OS with no distinction between RAM and HD ? Message-ID: Organization: Trax Softworks X-Newsreader: TIN [version 1.2 PL2] References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> Date: Fri, 10 Apr 1998 02:27:11 GMT Lines: 22 Sender: lfischer@netcom15.netcom.com Although Virtual Memory didn't pervade IBM's entire Mainframe line until the 1970s, it certainly had been "discovered" by IBM prior to 1968 - experimentally with the 360/40 modified for Virtual Memory hardware support and CP40/CMS in 1964, announced in 1965 as the 360/67 and TSS, and delivered to customers in 1967 as the 360/67 with either non-working TSS or working CP67/CMS. Most customers chose the working system. CP67 was the ancestor of VM/370. I am not of course claiming that IBM invented Virtual Memory. Len Richard M. Alderson III (alderson@netcom.netcom.com) wrote: : In article <6gijrn$f7o$1@flood.weeg.uiowa.edu> jones@pyrite.cs.uiowa.edu : (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: : >By the 1970's mainframe computer systems routinely used the virtual memory : >ideas that began with UNIX. : Virtual memory was already in use on PDP-10 systems in 1968, and was old tech- : nology even then. It simply hadn't been discovered by IBM as yet... ...snip... ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!masternews.telia.net!feed1.news.luth.se!luth.se!news.kth.se!grey04.nada.kth.se!d96-abo From: d96-abo@grey04.nada.kth.se (Alexander Boström) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 10 Apr 1998 09:38:09 GMT Lines: 28 Message-ID: <6gkpa1$21d$1@news.kth.se> References: <6gi2vo$gkm$1@news.kth.se> NNTP-Posting-Host: grey04.nada.kth.se Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Newsreader: TIN [version 1.2 PL2] Trevor Zion Bauknight (trev@sc.edu) wrote: : In article <6gi2vo$gkm$1@news.kth.se>, d96-abo@escapen.nada.kth.se : (Alexander Boström) wrote: : > You should have a look at the OS:es that gets rid of files, and instead : > just considers everything an object (a C "struct", a LISP "pair", a Java [...] : Now, I guess, FORTH is primarily used in imbedded situations and in : low-level stuff like Open Firmware because of the ease of porting it to : new architectures... : Is this what you mean? I'm no expert, but programming-wise, FORTH is the : only thing (other than Z80 assembly, reluctantly) that I ever spent any : time at all with... I've never used FORTH myself, but from what I gather it seems too low-level to be useful, however I'm probably wrong. It probably is an example of the kind of persistent-object system I was talking about, though. I have a friend who likes FORTH, and he thinks C is too low-level to be a general purpose programming language. See his language review page: http://www2.tunes.org/doc/Review/Languages.html#FORTH http://www2.tunes.org/doc/Review/Languages.html#C : Trev ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!nntp.abs.net!howland.erols.net!newsfeed.internetmci.com!24.128.1.125!chnws03.mediaone.net!24.131.1.12!denws01.mw.mediaone.net!news.gmi.edu!nova.kettering.edu!lee1089 From: Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Fri, 10 Apr 1998 15:16:18 -0400 Organization: Kettering University (formerly GMI E&MI) - Flint MI Lines: 16 Message-ID: References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> NNTP-Posting-Host: nova.kettering.edu Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII In-Reply-To: On 10 Apr 1998, John Hughes wrote: > jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: > > > Today, my understanding is that MAC OS and Windows 95 can be configured to > > do this too, > > You can't configure Win95 without VM. Yes, you can. Go into the System control panel. Under the fourth tab (Performance, I think) click Virtual Memory. Click "No Virtual Memory" or something like that. (Sorry but I'm working from memory without a W95 box in front of me.) [snip] ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news-fra1.dfn.de!news-was.dfn.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!newshub.northeast.verio.net!dca1-hub1.news.digex.net!digex!newsreader.digex.net!news1.chicago.iagnet.net!iagnet.net!news.uiowa.edu!not-for-mail From: jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 10 Apr 1998 15:36:14 GMT Organization: The University of Iowa Lines: 28 Message-ID: <6gle9e$lc8$1@flood.weeg.uiowa.edu> References: NNTP-Posting-Host: pyrite.cs.uiowa.edu From article , by John Hughes : > jones@pyrite.cs.uiowa.edu > (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: > >> By the 1970's mainframe computer systems routinely used the virtual memory >> ideas that began with UNIX. > > Revisionism of the highest order. Let me be more specific: From BSD UNIX we get the lazy implementation of fork, where the parent and child process share all pages of read write data until such time as one or the other process changes something. From BSD UNIX, we get the technology that gives us nearly LRU page replacement policies without the need to support even a mark bit in the MMU. It's all done in software, and this lets us have MMU's that conform to the RISC philosophy! Those ideas don't date back to the 1960's. They don't come from Multics, they're not from the Berkeley TSS on the SDS 940, and they're certainly not from Atlas. We do, however, owe debts to all these machines and their operating systems, and particularly to Belady at IBM research labs for his spectacular work on page replacement policies in the late 1960's. Doug Jones jones@cs.uiowa.edu ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news-peer.gip.net!news-raspail.gip.net!news.gsl.net!gip.net!oleane!calvacom!not-for-mail From: John Hughes Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 10 Apr 1998 16:29:09 +0200 Organization: Atlantic Technologies INC. Lines: 12 Message-ID: References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> NNTP-Posting-Host: microlite.calvacom.fr X-Newsreader: Gnus v5.5/XEmacs 20.3 - "Vatican City" jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: > By the 1970's mainframe computer systems routinely used the virtual memory > ideas that began with UNIX. Revisionism of the highest order. -- John Hughes , Atlantic Technologies Inc. Tel: +33-1-43204546 24 rue Montbrun, Fax: +33-1-43204579 75014 PARIS. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newscore.univie.ac.at!news-raspail.gip.net!news.gsl.net!gip.net!oleane!calvacom!not-for-mail From: John Hughes Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 10 Apr 1998 16:32:52 +0200 Organization: Atlantic Technologies INC. Lines: 19 Message-ID: References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> NNTP-Posting-Host: microlite.calvacom.fr X-Newsreader: Gnus v5.5/XEmacs 20.3 - "Vatican City" jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: > Today, my understanding is that MAC OS and Windows 95 can be configured to > do this too, You can't configure Win95 without VM. > ... Unlike Multics, though, these systems don't let you open files > by mapping the file into the memory address space. The last system > I know of that supported that beautiful idea was BSD 4.3 UNIX, and > hardly anyone ever used that feature. Um. Do you think BSD 4.4 can't do mmap? What about SVR4? Linux? -- John Hughes , Atlantic Technologies Inc. Tel: +33-1-43204546 24 rue Montbrun, Fax: +33-1-43204579 75014 PARIS. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!nntp.abs.net!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!rill.news.pipex.net!pipex!join.news.pipex.net!pipex!krypton.inbe.net!INbe.net!not-for-mail From: lucvdv@null.net (Luc Van der Veken) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Fri, 10 Apr 1998 19:07:14 GMT Organization: . Lines: 17 Message-ID: <35356c97.6249903@news.innet.be> References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> NNTP-Posting-Host: pool02a-194-7-46-227.uunet.be Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/32.451 X-No-Archive: yes John Hughes told us > jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: > > > Today, my understanding is that MAC OS and Windows 95 can be configured to > > do this too, > > You can't configure Win95 without VM. Yes you can (but you wouldn't want to - that's another matter). From within any program: Win-logo-key + Break / Rightarrow 3 times / Alt-V / M / S. (are they trying to be funny, making you type Alt-VMS at the end, or is it just a coincidence? ;) ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!nntp.abs.net!howland.erols.net!sunqbc.risq.qc.ca!news.uunet.ca!gts!bdb From: bdb@GTS.Net (Bruce Becker) Subject: Re: OS with no distinction between RAM and HD ? Message-ID: Organization: G.T.S., Toronto, Ontario X-Newsreader: trn 4.0-test63 (15 March 1998) References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> Date: Fri, 10 Apr 1998 19:56:57 GMT Lines: 22 In article <6gijrn$f7o$1@flood.weeg.uiowa.edu>, Douglas W. Jones,201H MLH,3193350740,3193382879 wrote: | |Today, my understanding is that MAC OS and Windows 95 can be configured to |do this too, although this is a recent development. Unlike Multics, though, |these systems don't let you open files by mapping the file into the memory |address space. The last system I know of that supported that beautiful |idea was BSD 4.3 UNIX, and hardly anyone ever used that feature. Sadly, my |antique UNIX system is running BSD 4.2, so I can't play with that. Most modern Unixes (Unices? 8^) support some form of facility to map files and/or devices into memory. Usually it's called "mmap()", but beware of brokenness & subtle incompatibilities in earlier versions... -- ,u, Bruce Becker Toronto, Ontario 1 416 699 1868 a /i/ Internet: bdb@gts.org Uucp: ...!gts!bdb `\o\-e "When did the Me Generation become the meme generation?" _< /_ - Mimi Pond, historian ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!nntp.news.xara.net!xara.net!supernews.xara.net!xara.net!SnUK10!baron.netcom.net.uk!netcom.net.uk!server3.netnews.ja.net!news.ox.ac.uk!not-for-mail From: Ben Hutchings Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 11 Apr 1998 03:47:19 +0100 Organization: Not organised Lines: 22 Sender: womble@eldritch.dyn.ml.org Message-ID: <6gmljn$86i$1@eldritch.dyn.ml.org> References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> NNTP-Posting-Host: max85.public.ox.ac.uk In article , wrote: >On 10 Apr 1998, John Hughes wrote: > >> jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: >> >> > Today, my understanding is that MAC OS and Windows 95 can be configured to >> > do this too, >> >> You can't configure Win95 without VM. > >Yes, you can. Go into the System control panel. Under the fourth tab >(Performance, I think) click Virtual Memory. Click "No Virtual Memory" or >something like that. (Sorry but I'm working from memory without a W95 box >in front of me.) But that actually disables the swap file, not virtual memory. -- Ben Hutchings | ICOA User Rep election 1998: http://www.jms.org/election/ email/finger: m95bwh@ecs.ox.ac.uk | web: http://users.ox.ac.uk/~worc0223/ friends: People who know you well, but like you anyway. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news.idt.net!newsfeed.nyu.edu!newsfeed.sgi.net!news.sgi.net!not-for-mail Message-ID: <352F056B.6CA0@NO-SPAM.sgi.net> From: Adam Stouffer X-Mailer: Mozilla 3.04 (WinNT; I) MIME-Version: 1.0 Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> <6gmljn$86i$1@eldritch.dyn.ml.org> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 14 Date: Sat, 11 Apr 1998 05:46:58 GMT NNTP-Posting-Host: dap-144-109.monroe.sgi.net NNTP-Posting-Date: Sat, 11 Apr 1998 01:46:58 EST Ben Hutchings wrote: > > > But that actually disables the swap file, not virtual memory. > What is your definition of virtual memory? -- _________________________ | | | Remove the NO-SPAM | | | _________________________ ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!nntp.abs.net!feed2.news.erols.com!erols!woodstock.news.demon.net!demon!news.demon.co.uk!demon!cucumber.demon.co.uk!usenet From: andrew@cucumber.demon.co.uk (Andrew Gabriel) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 11 Apr 1998 11:00:42 GMT Organization: home Message-ID: <6gnigq$2h2@cucumber.demon.co.uk> References: <35294515.1234@cs.tu-berlin.de> <6gijrn$f7o$1@flood.weeg.uiowa.edu> <352ef8ff.1461279@news.innet.be> <6gj3lr$557$1@eldritch.dyn.ml.org> <6gjdss$o7m$1@flood.weeg.uiowa.edu> NNTP-Posting-Host: cucumber X-NNTP-Posting-Host: cucumber.demon.co.uk:158.152.58.86 X-Trace: news.demon.co.uk 892293359 nnrp-07:13600 NO-IDENT cucumber.demon.co.uk:158.152.58.86 X-Complaints-To: abuse@demon.net X-Newsreader: knews 0.9.6 Lines: 16 In article <6gjdss$o7m$1@flood.weeg.uiowa.edu>, jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: >From article <6gj3lr$557$1@eldritch.dyn.ml.org>, by Ben Hutchings : > >> As do the UNIX and UNIX-alike systems I've used, which are not based >> on BSD. Linux actually loads executables by simply mapping them into >> a new process's memory. > >I find no evidence of this in the pure System V derivitives. Solaris does this, and I'm reasonably sure all other SVR4's do too. -- Andrew Gabriel Consultant Software Engineer ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed2.uk.ibm.net!sackheads.org!ibm.net!nntp.news.xara.net!xara.net!server5.netnews.ja.net!server3.netnews.ja.net!news.ox.ac.uk!not-for-mail From: Ben Hutchings Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 11 Apr 1998 15:16:58 +0100 Organization: Not organised Lines: 17 Sender: womble@max82.public.ox.ac.uk Message-ID: <6gnu0q$np8$1@max82.public.ox.ac.uk> References: <6gmljn$86i$1@eldritch.dyn.ml.org> <352F056B.6CA0@NO-SPAM.sgi.net> NNTP-Posting-Host: max82.public.ox.ac.uk In article <352F056B.6CA0@NO-SPAM.sgi.net>, Adam Stouffer wrote: >Ben Hutchings wrote: >> >> >> But that actually disables the swap file, not virtual memory. >> > >What is your definition of virtual memory? Memory being accessed through virtual addresses which, when used, are automatically translated to physical addresses or else flagged invalid. -- Ben Hutchings | ICOA User Rep election 1998: http://www.jms.org/election/ email/finger: m95bwh@ecs.ox.ac.uk | web: http://users.ox.ac.uk/~worc0223/ Humour is the best antidote to reality. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed1.earthlink.net!nntp.earthlink.net!usenet From: "Joel C. Ewing" Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sat, 11 Apr 1998 18:24:47 -0400 Organization: EarthLink Network, Inc. Lines: 40 Message-ID: <6gou4r$s5q@chile.earthlink.net> References: <6gmljn$86i$1@eldritch.dyn.ml.org> <352F056B.6CA0@NO-SPAM.sgi.net> <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> Reply-To: "Joel C. Ewing" NNTP-Posting-Host: 208.254.205.137 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 2.02 (OS/2; I) Adam Stouffer wrote: > > Ben Hutchings wrote: > > Memory being accessed through virtual addresses which, when used, are > > automatically translated to physical addresses or else flagged invalid. > > Ok, so virtual memory and a swap file are different? > The usual connotation of "swap file" is that all memory associated with a suspended task is moved to auxiliary storage, and for the task to resume execution, all its storage must first be restored to real memory. "Virtual memory" implies that a running task (thread, whatever the name) sees memory as one (or possibly more) linear address spaces which appear to the running program as if they reside in real memory, but in reality those pieces (pages) of the linear storage which haven't been referenced for a long time may have been migrated to auxiliary storage (by the operating system, to make room for other pages that were needed). The distinction is that with virtual memory a task may make execution progress while only a part of its storage space (the "working set") is in real memory, while with just swapping, execution progress requires the entire storage space to be in real memory. Swapping allows the sum of the memory requirements of all active threads in the operating system to exceed to actual real memory, but each individual thread must be able to execute within the constraints of real memory. Virtual memory allows the memory requirements of a single active thread to exceed real memory capacity, and as long as the active working set can be contained in real memory, reasonable performance can be sustained. I don't know the internals of current PC Operating Systems well enough to know for sure whether they use the term "swap file" in the classical sense, but I suspect that if it is possible for an application running under these systems to exceed available real memory, that they can only do it by performing special system requests to swap or overlay segments of memory not currently needed by the application. If that is indeed the case, then these are NOT examples of true virtual memory. -- Joel C. Ewing, Fort Smith, AR jcewing@acm.org ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed1.earthlink.net!nntp.earthlink.net!usenet From: "Joel C. Ewing" Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sat, 11 Apr 1998 18:24:47 -0400 Organization: EarthLink Network, Inc. Lines: 40 Message-ID: <6gou4r$s5q@chile.earthlink.net> References: <6gmljn$86i$1@eldritch.dyn.ml.org> <352F056B.6CA0@NO-SPAM.sgi.net> <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> Reply-To: "Joel C. Ewing" NNTP-Posting-Host: 208.254.205.137 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 2.02 (OS/2; I) Adam Stouffer wrote: > > Ben Hutchings wrote: > > Memory being accessed through virtual addresses which, when used, are > > automatically translated to physical addresses or else flagged invalid. > > Ok, so virtual memory and a swap file are different? > The usual connotation of "swap file" is that all memory associated with a suspended task is moved to auxiliary storage, and for the task to resume execution, all its storage must first be restored to real memory. "Virtual memory" implies that a running task (thread, whatever the name) sees memory as one (or possibly more) linear address spaces which appear to the running program as if they reside in real memory, but in reality those pieces (pages) of the linear storage which haven't been referenced for a long time may have been migrated to auxiliary storage (by the operating system, to make room for other pages that were needed). The distinction is that with virtual memory a task may make execution progress while only a part of its storage space (the "working set") is in real memory, while with just swapping, execution progress requires the entire storage space to be in real memory. Swapping allows the sum of the memory requirements of all active threads in the operating system to exceed to actual real memory, but each individual thread must be able to execute within the constraints of real memory. Virtual memory allows the memory requirements of a single active thread to exceed real memory capacity, and as long as the active working set can be contained in real memory, reasonable performance can be sustained. I don't know the internals of current PC Operating Systems well enough to know for sure whether they use the term "swap file" in the classical sense, but I suspect that if it is possible for an application running under these systems to exceed available real memory, that they can only do it by performing special system requests to swap or overlay segments of memory not currently needed by the application. If that is indeed the case, then these are NOT examples of true virtual memory. -- Joel C. Ewing, Fort Smith, AR jcewing@acm.org ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news-fra1.dfn.de!bignews.mediaways.net!fu-berlin.de!news-peer.gip.net!news.gsl.net!gip.net!news-peer.sprintlink.net!news-backup-east.sprintlink.net!news-in-east.sprintlink.net!news.sprintlink.net!Sprint!205.237.233.10!wesley.videotron.net!newsfeed.nyu.edu!newsfeed.sgi.net!news.sgi.net!not-for-mail Message-ID: <352FD629.7565@NO-SPAM.sgi.net> From: Adam Stouffer X-Mailer: Mozilla 3.04 (WinNT; I) MIME-Version: 1.0 Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? References: <6gmljn$86i$1@eldritch.dyn.ml.org> <352F056B.6CA0@NO-SPAM.sgi.net> <6gnu0q$np8$1@max82.public.ox.ac.uk> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 16 Date: Sat, 11 Apr 1998 20:37:36 GMT NNTP-Posting-Host: dap-144-99.monroe.sgi.net NNTP-Posting-Date: Sat, 11 Apr 1998 16:37:36 EST Ben Hutchings wrote: > > > Memory being accessed through virtual addresses which, when used, are > automatically translated to physical addresses or else flagged invalid. Ok, so virtual memory and a swap file are different? Adam -- _________________________ | | | Remove the NO-SPAM | | | _________________________ ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed2.uk.ibm.net!sackheads.org!ibm.net!news-lond.gip.net!news-peer.gip.net!news.gsl.net!gip.net!news-peer.sprintlink.net!news-peer-west.sprintlink.net!news.sprintlink.net!Sprint!newsfeed.wli.net!newshub.sirius.com!newsfiler.sirius.com!hnsngr From: hnsngr@sirius.com (Ron Hunsinger) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sun, 12 Apr 1998 03:37:36 -0700 Organization: ErsteSoft Lines: 35 Message-ID: References: <6gmljn$86i$1@eldritch.dyn.ml.org> <352F056B.6CA0@NO-SPAM.sgi.net> <6gnu0q$np8$1@max82.public.ox.ac.uk> NNTP-Posting-Host: ppp-asft01--026.sirius.net Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit X-Newsreader: Yet Another NewsWatcher 2.3.1 In article <6gnu0q$np8$1@max82.public.ox.ac.uk>, Ben Hutchings wrote: > Memory being accessed through virtual addresses which, when used, are > automatically translated to physical addresses or else flagged invalid. That's address translation. While address translation is a prerequisite of virtual memory, it's pushing it more than a little to equate the concepts. To call it virtual memory, you need at least that the mapping is dynamic. In fact, I'd be reluctant to call it virtual memory unless one of the normal responses to detecting an invalid address was to somehow make the address valid and restart the failed instruction. As a case in point, 68030-based Macintoshes will let you turn virtual memory on or off, but even when it's turned off the address translation provided by the PMMU (built into the 68030) is active. It does two things: Physically, all memory banks are the same size, so unless the bank is populated with RAM of the maximum size there will be a gap in the physical addresses between them. The address translation is used to hide that gap, making it appear that RAM is contiguous. On models that use RAM as VRAM, the video memory is physically at address 0 (where it's easy for the built-in video refresh circuitry to find it), but the address translation makes it appear to the processor to be elsewhere (because the processor expects the trap vector table to begin at address 0). With virtual memory turned off, the address translation table is set up at initialization time, and never changed thereafter. It's still active, and useful (reducing the number of auxiliary chips needed), but it isn't virtual memory. -Ron Hunsinger ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news.he.net!Supernews73!supernews.com!Supernews69!not-for-mail From: genew@vip.net (Gene Wirchenko) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sun, 12 Apr 1998 06:01:08 GMT Organization: All USENET -- http://www.Supernews.com Lines: 24 Message-ID: <352efe05.3116599@news.vip.net> References: <6gijrn$f7o$1@flood.weeg.uiowa.edu> Reply-To: genew@vip.net NNTP-Posting-Host: 1688@204.209.212.50 X-Newsreader: Forte Free Agent 1.1/32.230 John Hughes wrote: >jones@pyrite.cs.uiowa.edu (Douglas W. Jones,201H MLH,3193350740,3193382879) writes: > >> Today, my understanding is that MAC OS and Windows 95 can be configured to >> do this too, > >You can't configure Win95 without VM. Sure you can: Start - Settings - Control Panel - System - Performance - Virtual Memory - Let me specify my own virtual memory settings. - Disable virtual memory (not recommended). - OK - OK. [snip] Sincerely, Gene Wirchenko C Pronunciation Guide: y=x++; "wye equals ex plus plus semicolon" x=x++; "ex equals ex doublecross semicolon" ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!nntp.news.xara.net!xara.net!server5.netnews.ja.net!server3.netnews.ja.net!news.ox.ac.uk!not-for-mail From: Ben Hutchings Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 12 Apr 1998 08:29:28 +0100 Organization: Not organised Lines: 52 Sender: womble@eldritch.dyn.ml.org Message-ID: <6gpqgp$468$1@eldritch.dyn.ml.org> References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> NNTP-Posting-Host: max51.public.ox.ac.uk In article <6gou4r$s5q@chile.earthlink.net>, Joel C. Ewing wrote: >Adam Stouffer wrote: >> >> Ben Hutchings wrote: >> > Memory being accessed through virtual addresses which, when used, are >> > automatically translated to physical addresses or else flagged invalid. >> >> Ok, so virtual memory and a swap file are different? >> > >The usual connotation of "swap file" is that all memory associated with >a suspended task is moved to auxiliary storage, and for the task to >resume execution, all its storage must first be restored to real >memory. Not to me. I think it can also be a file to which pages of memory are swapped. This is certainly the terminology that Linux uses. >"Virtual memory" implies that a running task (thread, whatever the >name) sees memory as one (or possibly more) linear address spaces >which appear to the running program as if they reside in real memory, >but in reality those pieces (pages) of the linear storage which >haven't been referenced for a long time may have been migrated to >auxiliary storage (by the operating system, to make room for other >pages that were needed). But not only that, some pages of virtual memory may be invalid and cause segment violations when accessed. So virtual memory implies that there is a translation involved - and this can be used to facilitate not only swapping but memory protection, demand loading, stack extension, etc. >I don't know the internals of current PC Operating Systems well enough >to know for sure whether they use the term "swap file" in the classical >sense, but I suspect that if it is possible for an application running >under these systems to exceed available real memory, that they can only >do it by performing special system requests to swap or overlay segments >of memory not currently needed by the application. If that is indeed >the case, then these are NOT examples of true virtual memory. The i386 architecture provides for paged segmented virtual memory, and operating systems for it (Windows 3.x, 95, NT, Linux, *BSD) *do* use that to provide swapping on a page-by-page basis, among other things. Applications do *not* specifically request this. -- Ben Hutchings | ICOA User Rep election 1998: http://www.jms.org/election/ email/finger: m95bwh@ecs.ox.ac.uk | web: http://users.ox.ac.uk/~worc0223/ Death is a nonmaskable interrupt. ###### Newsgroups: alt.folklore.computers Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news.gsl.net!gip.net!news.he.net!Supernews60!supernews.com!coop.net!world!dpbsmith From: dpbsmith@world.std.com (Daniel P. B. Smith) Subject: Re: OS with no distinction between RAM and HD ? Message-ID: Organization: The World Public Access UNIX, Brookline, MA References: <6g8i2f$cga$1@news1.rmi.net> Date: Mon, 13 Apr 1998 02:11:35 GMT Lines: 46 In article <6g8i2f$cga$1@news1.rmi.net>, Andy Rabagliati wrote: >According to Simon Slavin : >> >> Was there ever an OS that just dealt with disk as if it's >> part of main memory ? I don't know whether you would call M (MUMPS) an OS or not; nor do I know exactly what criteria you would set for "just dealing with disk as if it's part of main memory." However... In M, there are fundamentally no such things as files. Basically, M has one universal data type (it acts mostly like a string, with automatic conversions to and from integers and floats as driven by the context) and one universal data structure. The universal data structure is an array, except that it has an indefinite number of dimensions, an indefinite number of index values for each dimension, and the index values are general M variables and hence can assume integer and, most usefully, string values. Everything is dynamically assigned as needed and stored sparsely in B-trees, and because this is so fundamental to the way the language works, market pressure force implementors to do this in efficient ways. So in M, you simply set ^ssn("smith","dan")="314-15-2653" and, presto, variable ssn, sub "smith", sub "dan" has been set equal to Dan Smith's social security number. Now, if the variable name does not begin with that caret, the variable is a local variable, and is stored in RAM, "belongs" to the process that created it, and is volatile (vanishes when the process terminates) On the other hand, if it does begin with a caret, the variable is global, can be accessed by all processes, is stored on the disk, and is persistent. In real life, real, large databases (gigabytes) can and are stored in this way. Disk files can be accessed, but are only used for things like sequential I/O of text reports to be output, or for communication between M programs and file-oriented programs. This general design has been part of M since at least the late seventies. -- Daniel P. B. Smith dpbsmith@world.std.com ###### Newsgroups: alt.folklore.computers Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news.gsl.net!gip.net!news.he.net!Supernews60!supernews.com!coop.net!world!dpbsmith From: dpbsmith@world.std.com (Daniel P. B. Smith) Subject: Re: OS with no distinction between RAM and HD ? Message-ID: Organization: The World Public Access UNIX, Brookline, MA References: <6g8i2f$cga$1@news1.rmi.net> Date: Mon, 13 Apr 1998 02:11:35 GMT Lines: 46 In article <6g8i2f$cga$1@news1.rmi.net>, Andy Rabagliati wrote: >According to Simon Slavin : >> >> Was there ever an OS that just dealt with disk as if it's >> part of main memory ? I don't know whether you would call M (MUMPS) an OS or not; nor do I know exactly what criteria you would set for "just dealing with disk as if it's part of main memory." However... In M, there are fundamentally no such things as files. Basically, M has one universal data type (it acts mostly like a string, with automatic conversions to and from integers and floats as driven by the context) and one universal data structure. The universal data structure is an array, except that it has an indefinite number of dimensions, an indefinite number of index values for each dimension, and the index values are general M variables and hence can assume integer and, most usefully, string values. Everything is dynamically assigned as needed and stored sparsely in B-trees, and because this is so fundamental to the way the language works, market pressure force implementors to do this in efficient ways. So in M, you simply set ^ssn("smith","dan")="314-15-2653" and, presto, variable ssn, sub "smith", sub "dan" has been set equal to Dan Smith's social security number. Now, if the variable name does not begin with that caret, the variable is a local variable, and is stored in RAM, "belongs" to the process that created it, and is volatile (vanishes when the process terminates) On the other hand, if it does begin with a caret, the variable is global, can be accessed by all processes, is stored on the disk, and is persistent. In real life, real, large databases (gigabytes) can and are stored in this way. Disk files can be accessed, but are only used for things like sequential I/O of text reports to be output, or for communication between M programs and file-oriented programs. This general design has been part of M since at least the late seventies. -- Daniel P. B. Smith dpbsmith@world.std.com ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!news-nyc.telia.net!masternews.telia.net!news-feed.inet.tele.dk!bofh.vszbr.cz!enews.sgi.com!decwrl!tribune.usask.ca!news.uregina.ca!not-for-mail From: bayko@borealis.cs.uregina.ca (John Bayko) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 13 Apr 1998 16:07:12 GMT Organization: University of Regina, Dept. of Computer Science Lines: 28 Message-ID: <6gtd7g$10s6$1@sue.cc.uregina.ca> References: <6gmljn$86i$1@eldritch.dyn.ml.org> <352F056B.6CA0@NO-SPAM.sgi.net> <6gnu0q$np8$1@max82.public.ox.ac.uk> NNTP-Posting-Host: borealis.cs.uregina.ca In article <6gnu0q$np8$1@max82.public.ox.ac.uk>, Ben Hutchings wrote: >In article <352F056B.6CA0@NO-SPAM.sgi.net>, >Adam Stouffer wrote: >> >>What is your definition of virtual memory? > >Memory being accessed through virtual addresses which, when used, are >automatically translated to physical addresses or else flagged invalid. That is virtual addressing, the memory which it addresses remains real. 'Virtual', properly used, refers to something that acts like one thing, but is implemented by a substitute mechanism. Virtual memory refers to internal memory which is not real, but acts as if it were, normally implemented by some sort of direct access storage, like a HD. It's entirely possible to have virtual memory without virtual addressing (such as in a capability-based operating system perhaps?). What you call 'virtual addresses' are more properly called 'logical addresses'. Similarly, you can support 'virtual connections' over packet-based networks (TCP/IP over Ethernet) which don't support actual physical or logical connections, or multiple 'vurtual connections' over a single physical connection (PPP) which act as a substitute for real connections. -- John Bayko (Tau). bayko@cs.uregina.ca http://www.cs.uregina.ca/~bayko ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed2.uk.ibm.net!sackheads.org!ibm.net!news-lond.gip.net!news.gsl.net!gip.net!nntp.news.xara.net!xara.net!server5.netnews.ja.net!server3.netnews.ja.net!news.ox.ac.uk!not-for-mail From: Ben Hutchings Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 13 Apr 1998 18:42:36 +0100 Organization: Not organised Lines: 27 Sender: womble@eldritch.dyn.ml.org Message-ID: <6gtiqc$1g6$1@eldritch.dyn.ml.org> References: <352F056B.6CA0@NO-SPAM.sgi.net> <6gnu0q$np8$1@max82.public.ox.ac.uk> <6gtd7g$10s6$1@sue.cc.uregina.ca> NNTP-Posting-Host: max66.public.ox.ac.uk In article <6gtd7g$10s6$1@sue.cc.uregina.ca>, John Bayko wrote: >In article <6gnu0q$np8$1@max82.public.ox.ac.uk>, > Ben Hutchings wrote: >>In article <352F056B.6CA0@NO-SPAM.sgi.net>, >>Adam Stouffer wrote: >>> >>>What is your definition of virtual memory? >> >>Memory being accessed through virtual addresses which, when used, are >>automatically translated to physical addresses or else flagged invalid. > > That is virtual addressing, the memory which it addresses remains >real. 'Virtual', properly used, refers to something that acts like one >thing, but is implemented by a substitute mechanism. Virtual memory >refers to internal memory which is not real, but acts as if it were, >normally implemented by some sort of direct access storage, like a HD. Well, that's what you can do by flagging certain virtual addresses as invalid and using an exception handler to deal with them. If addresses are merely translated, I'd agree that's not a virtual memory system. -- Ben Hutchings -- should be revising | Jay Miner Society: http://www.jms.org/ m95bwh@ecs.ox.ac.uk | http://users.ox.ac.uk/~worc0223/ | IRC: Womble In a hierarchy, every employee tends to rise to his level of incompetence. ###### Path: ccw.ch!usenet From: Neil Franklin Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Mon, 13 Apr 1998 22:24:49 +0200 Organization: My own Private Self Lines: 61 Message-ID: <35327491.E93FA0ED@ccw.ch> References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> <6gpqgp$468$1@eldritch.dyn.ml.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.04 [en] (X11; I; Linux 2.0.27 i486) Ben Hutchings wrote: > Joel C. Ewing wrote: > >Adam Stouffer wrote: > >> Ben Hutchings wrote: > >> > Memory being accessed through virtual addresses which, when used, are > >> > automatically translated to physical addresses or else flagged invalid. > >> Ok, so virtual memory and a swap file are different? Yes, different. VM does not require an swapfile (actually an pagefile). Simply mapping logical address space to physical RAM and doing demand (re-)loading of executables fullfils the requirement for VM. There is no need to page-out long-not-accessed memory that did not come from an file to an swap/pagefile. That just frees its pages to enlarge the paging pool, and so reduces the amount of page-ins. In Linux simply type "swapoff" and you have got VM without an swap/pagefile (which you may delete, but elliminate the mount line from fstab first). My notebook constanly runs like this (lots of RAM and a little disk). > >The usual connotation of "swap file" is that all memory associated with > >a suspended task is moved to auxiliary storage, and for the task to > >resume execution, all its storage must first be restored to real > Not to me. I think it can also be a file to which pages of memory are > swapped. This is certainly the terminology that Linux uses. What Linux (and *BDS) does is in CS called VM with paging. But is has been a tradition in Unix, since paging was introduced, to call it swapping (Unix had swapping of processes since day 1). > But not only that, some pages of virtual memory may be invalid and > cause segment violations when accessed. So virtual memory implies > that there is a translation involved Translation is needed, but not an swap/pagefile. That is only needed if you want to page-out to increase page pool size. > >I don't know the internals of current PC Operating Systems well enough > >to know for sure whether they use the term "swap file" in the classical > The i386 architecture provides for paged segmented virtual memory, and > operating systems for it (Windows 3.x, 95, NT, Linux, *BSD) *do* use > that to provide swapping on a page-by-page basis, among other things. > Applications do *not* specifically request this. All PC OSes I know (exept DOS) use full VM, paging and a pagefile. The option in Windows mentioned earlier (at least in W3.x) switched off using an pagefile. VM and demand loading kept on running. -- private: Neil.Franklin.remove.this@ccw.ch http://www.ccw.ch/Neil.Franklin/ office: franklin.remove.this@arch.ethz.ch http://caad.arch.ethz.ch/~franklin/ Can a Microsoft allergy be claimed on job health insurence? ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsxfer3.itd.umich.edu!rill.news.pipex.net!pipex!join.news.pipex.net!pipex!krypton.inbe.net!INbe.net!not-for-mail From: lucvdv@null.net (Luc Van der Veken) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Tue, 14 Apr 1998 20:06:30 GMT Organization: . Lines: 68 Message-ID: <3533a21c.1740195@news.innet.be> References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> <6gpqgp$468$1@eldritch.dyn.ml.org> <35327491.E93FA0ED@ccw.ch> NNTP-Posting-Host: pmpool053-52.innet.be Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/32.451 X-No-Archive: yes Neil Franklin got near to the point: > All PC OSes I know (exept DOS) use full VM, paging and a pagefile. The > option in Windows mentioned earlier (at least in W3.x) switched off > using an pagefile. Someone got it right, at last. (I know I pointed out myself how to disable it in response to someone saying you couldn't - at that time I thought he was referring to VM for paging, I wanted to keep it short 'n simple and hadn't expected this thread to grow so long.) > VM and demand loading kept on running. VM: OK, but if there's no swapfile (or pagefile, if you prefer that name) to load demanded pages from or store them into, how is demand loading going to keep on running? As I always understood VM (repeat: as *I* see it), virtual means "not really there", and memory in this context means the processor's working memory, i.e. RAM. Have 16 megs, can address 32 megs and find different data everywhere. Any (hardware) system that makes it look as if you have more RAM than there really is, provides VM. (and any system that makes it look as if there's LESS than there really is, must be provided by MS - or no, EMACS isn't, and that already was Eight Megs And Constantly Swapping some 10 versions ago ;) Back to the subject now. If you can map a file (*any* file from disk, tape, papertape or whatever, just as long as it isn't in RAM) to an unused portion of your memory space and access it just as if its contents were in RAM, that's VM. If a page/swap file is used to free up some RAM, the easiest way to do it (supposing you have mapping and trapping capabilities) is to map a swap file (win) or a swap partition (Linux) into some unused address range, and move less used parts of RAM into that address range, freeing up 'real' memory. That's using VM for paging. Now you can use trapping to detect an attempted access to that RAM at its old location, swap it back in (using simple MOV instructions if you want to, since it's in a memory mapped file), and translate the address where the user thought he would find it to the new real address. That last thing is address translation: can exist /beside/ VM, but not necessarily, and has not much else to do with it - except that they usually come together because one is easier to add in hardware if the other is already there. If you don't have the above capabilities, you can *still* swap or page. The best example is a utility for DOS, that was once described (with full source) in Dr Dobbs Journal: even on a 8086, it could swap almost everything that was in memory out to a disk file to make room for a child process, and swap it back in later. NO VM USED (unless you would want to call it software VM emulation, to make things even more complicated). Using VM is an easy way to implement paging/swapping, but it is not a necessity, let alone that they are the same like some people seem to think (and I must admit that at least in this thread I helped throw a little wood on that fire). ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newspump.monmouth.com!newspeer.monmouth.com!newsfeed.direct.ca!news.uunet.ca!atbowler From: atbowler@thinkage.on.ca (Alan Bowler) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 14 Apr 1998 20:09:40 GMT Organization: Thinkage Ltd. Lines: 13 Message-ID: <6h0fq4$813$1@nntp1.uunet.ca> References: <6gmljn$86i$1@eldritch.dyn.ml.org> <352F056B.6CA0@NO-SPAM.sgi.net> <6gnu0q$np8$1@max82.public.ox.ac.uk> NNTP-Posting-Host: 192.102.11.4 In article <6gnu0q$np8$1@max82.public.ox.ac.uk> Ben Hutchings writes: >In article <352F056B.6CA0@NO-SPAM.sgi.net>, >Adam Stouffer wrote: >> >>What is your definition of virtual memory? > >Memory being accessed through virtual addresses which, when used, are >automatically translated to physical addresses or else flagged invalid. This is the definition I use. People however get upset when I point out that this does not mention paging and includes the simple base and bounds relocation schemes used by processors like the GE-635 and the PDP-10 (before paging hardare was added.) ###### Path: ccw.ch!usenet From: Neil Franklin Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Wed, 15 Apr 1998 01:36:23 +0200 Organization: My own Private Self Lines: 137 Message-ID: <3533F2F7.59E2A36D@ccw.ch> References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> <6gpqgp$468$1@eldritch.dyn.ml.org> <35327491.E93FA0ED@ccw.ch> <3533a21c.1740195@news.innet.be> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.04 [en] (X11; I; Linux 2.0.27 i486) Luc Van der Veken wrote: > > Neil Franklin got near to the > point: > > >The option in Windows mentioned earlier (at least in W3.x) > > switched off using an pagefile. > > and hadn't expected this thread to grow so long.) Neither did I. That is why I waited so long to post into it. And here comes Computer Science 101 :-) > > VM and demand loading kept on running. > > VM: OK, but if there's no swapfile (or pagefile, if you prefer > that name) to load demanded pages from or store them into, how is > demand loading going to keep on running? Strategy one: by using the VM system only for loading pages from the executable when they are first used but then keeping all of them in memory until process exit (we could call this lazy loading in analogy to lazy evaluation). Strategy two: by having the VM system respond to full memory commitment by dropping pages, but _only_ from those pages that can be reloaded from an executable file (= are already on disk), not those that were generated or altered while run time, such as stacks, malloc()ed stuff, etc (= not already on disk). The later (called dirty pages) become fixed/locked in memory until process exit. In this view of things the swap/page file is simply there to save copies of dirty pages to disk (page-out), and so make them reloadable like pages that come from executables, and so make them droppable (clean). Note that the pageing out (cleaning) activity is run as an background thread, not as part of demand page-in. As I said page-out simply leads to an enlarged paging pool (all the clean pages, that can be potentially dropped to make place for page-in). Linux uses this second strategy. This is why swapoff only costs speed and why swap/pagefiles are not much slower than swap/pagepartitions. 4.3BSD used only the first strategy, with anything being dropped (clean or dirty) going to the swap/pagefile, reload only from there. This is due to the hardware/performance differences of 1980s VAXen and 1990s PCs. Which strategy modern *BSDs use I don't know. > Any (hardware) system that makes it look as if you have more RAM > than there really is, provides VM. Yes. Actually to be precise: makes it that you have more valid address range than RAM address range. Malloc() allocates address range, not RAM space on VM systems. > If you can map a file (*any* file from disk, tape, papertape or > whatever, just as long as it isn't in RAM) to an unused portion > of your memory space and access it just as if its contents were > in RAM, that's VM. No requirement for it to not be in RAM. mmap() simply declares an equvalence of an address range with an section of an file. Every process has an mapping table which splits its address range into blocks and equvalises each block with an an section of an file, executable or swap/pagefile. > If a page/swap file is used to free up some RAM, the easiest way > to do it (supposing you have mapping and trapping capabilities) > is to map a swap file (win) or a swap partition (Linux) into some > unused address range, and move less used parts of RAM into that > address range, freeing up 'real' memory. That's using VM for > paging. Actually that is an inefficient algorithm (it requires separate VM and mapping RAM pools and RAM to RAM copying). In Linux there is an second table that corresponds each RAM page to an disk page. Demand loading works by assigning a RAM page to a disk section, loading from disk into it and then mapping _that_ RAM page into the processes address space. It stays there immobile until RAM is short, then it is unmapped from its process and reused for an other assignment. The entire RAM (except that used by the kernal) is used as one big disk cache (page pool). Without a swap/pagefile the dirty pages are subtracted from the pool. The "used" and "-buffers" fields of the Linux "free" command simply report how much is mapped and how much of that is swap/pagefile mapped (or fixed). > Now you can use trapping to detect an attempted access to that > RAM at its old location, swap it back in (using simple MOV > instructions if you want to, since it's in a memory mapped file), In Linux you only need to trap and reload when a page was dropped from RAM to make room for an other page, so there is allways a page reload from disk. That is why DMA capable disk controllers add so much speed to an Linux system. > If you don't have the above capabilities, you can *still* swap or > page. The best example is a utility for DOS, that was once > described (with full source) in Dr Dobbs Journal: even on a 8086, > it could swap almost everything that was in memory out to a disk > file to make room for a child process, and swap it back in later. Actually you can with the proper software writing rules (calls and accesses to malloc()ed memory and perhaps even stack expansion via an trampoline table) implement demand loading on any processor ever made (assuming it has an block addressabel mass storage (disk, drum or tape attached). They already did this on the Whirlwind in the 1950s. The pre-System-7 Macs and Win 1.0 both had an not particularly good implementation of this in them (malloc()ed memory required explicit page/segment-in checking by the programmer, very bug attracting). Hey I am back on charter for a.f.c! > NO VM USED (unless you would want to call it software VM > emulation, to make things even more complicated). This _is_ called software VM. Proper would be to say no VM hardware used. -- private: Neil.Franklin.remove.this@ccw.ch http://www.ccw.ch/Neil.Franklin/ office: franklin.remove.this@arch.ethz.ch http://caad.arch.ethz.ch/~franklin/ Can a Microsoft allergy be claimed on job health insurence? ###### Newsgroups: alt.folklore.computers Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!uunet!in2.uu.net!xyzzy!nntp From: "John D. Burleson" Subject: Re: OS with no distinction between RAM and HD ? X-Nntp-Posting-Host: help.stl.mo.boeing.com Content-Type: text/plain; charset=us-ascii Message-ID: <3536410A.9A6D163B@boeing.com> Sender: nntp@news.boeing.com (Boeing NNTP News Access) Content-Transfer-Encoding: 7bit Organization: The Boeing Company - McDonnell Douglas - Huntsville Division References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> <6gpqgp$468$1@eldritch.dyn.ml.org> Mime-Version: 1.0 Date: Thu, 16 Apr 1998 17:34:02 GMT X-Mailer: Mozilla 4.05 [en] (X11; U; SunOS 5.5.1 sun4u) Lines: 47 Ben Hutchings wrote: > > In article <6gou4r$s5q@chile.earthlink.net>, > Joel C. Ewing wrote: > >Adam Stouffer wrote: > >> > >> Ben Hutchings wrote: > >> > Memory being accessed through virtual addresses which, when used, are > >> > automatically translated to physical addresses or else flagged invalid. > >> > >> Ok, so virtual memory and a swap file are different? > >> > > > >The usual connotation of "swap file" is that all memory associated with > >a suspended task is moved to auxiliary storage, and for the task to > >resume execution, all its storage must first be restored to real > >memory. > > > Not to me. I think it can also be a file to which pages of memory are > swapped. This is certainly the terminology that Linux uses. > Properly speaking, this is paging and swapping means moving the whole process as described. As various Unices began to add paged VM systems , the swap partition was used as paging space. Therefore the name was retained for all the wrong reasons and on modern UNIX variants (since the early 80's, I would guess) most folks seem to use the terms interchangeably. As for swapping (in the original sense) this was the original purpose of the "sticky bit" in the mode bits of UNIX file inodes. Applications with the sticky bit turned on stayed in the swap partition after execution (as it was, presumably, faster storage) so as to facilitate faster future invocations. It was very common for vi to have the sticky bit set. Of course over use of the sticky bit meant you could consume most of your swap space even if nothing was running (I rememeber a v6 kernel and a newbie admin... wait a minute, that was me!). Modern Unices use this bit to resolve some of the permission problems inherent in globally writable tmp directories. -- John Burleson (mailto:john.d.burleson@boeing.com) Principal Software Engineer The Boeing Company/McDonnell Douglas Aerospace/Huntsville (205)922-7589 FAX:(205)922-4890 ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newspump.monmouth.com!newspeer.monmouth.com!news.idt.net!news.maxwell.syr.edu!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!btnet-peer!btnet-feed1!btnet!peer.news.zetnet.net!zetnet.co.uk!not-for-mail From: lisard@zetnet.co.uk Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 18 Apr 1998 12:15:30 GMT Lines: 17 Message-ID: <6ha5h2$9q$1@irk.zetnet.co.uk> References: <3533F2F7.59E2A36D@ccw.ch> NNTP-Posting-Host: man-153.dialup.zetnet.co.uk X-Everything: Net-Tamer V 1.08X On 1998-04-15 Neil.Franklin.remove.this@ccw.ch said: :> Any (hardware) system that makes it look as if you have more RAM :> than there really is, provides VM. :Yes. Actually to be precise: makes it that you have more valid :address range than RAM address range. Malloc() allocates address :range, not RAM space on VM systems. what about the situation where you're using address remapping to pretend there's less memory than there actually is? (eg. to fit a 22-bit memory space into a 16-bit virtual address space) all the dec literature called this virtual memory, didn't it? -- Communa (together) we remember... we'll see you falling you know soft spoken changes nothing to sing within her... ###### Path: ccw.ch!usenet From: Neil Franklin Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sun, 19 Apr 1998 02:15:10 +0200 Organization: My own Private Self Lines: 38 Message-ID: <3539420E.C57D3DD2@ccw.ch> References: <3533F2F7.59E2A36D@ccw.ch> <6ha5h2$9q$1@irk.zetnet.co.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.04 [en] (X11; I; Linux 2.0.27 i486) lisard@zetnet.co.uk wrote: > > On 1998-04-15 Neil.Franklin.remove.this@ccw.ch said: > :> Any (hardware) system that makes it look as if you have more RAM > :> than there really is, provides VM. > > :Yes. Actually to be precise: makes it that you have more valid > :address range than RAM address range. Malloc() allocates address > :range, not RAM space on VM systems. > > what about the situation where you're using address remapping to pretend > there's less memory than there actually is? (eg. to fit a 22-bit memory > space into a 16-bit virtual address space) all the dec literature called > this virtual memory, didn't it? Eaarrmmmm. You got me on that one :-) I am going to try this escape: On an pdp11 you have multiple processes of each 16 bit address range. If you consider the process ID part of an "extended address" you still have mapping downwards. That is, assuming enough processes. Else you would have not thrown up the money for the RAM to use 22 bit addresses. With more RAM than address space being used you essentially would be not using VM, even if the architecture did support its use. Without this escape an 8086 with its 16->20 bit stuff would be VM. And I definitely do not want to give it that status :-) Actually I can offer annother point: the pdp11 had demand paging (8k pages IIRC), the 8086 doesn't, so that is what made it VM capable. -- private: Neil.Franklin.remove.this@ccw.ch http://www.ccw.ch/Neil.Franklin/ office: franklin.remove.this@arch.ethz.ch http://caad.arch.ethz.ch/~franklin/ Can a Microsoft allergy be claimed on job health insurence? ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newsgate.cistron.nl!het.net!newspump.monmouth.com!newspeer.monmouth.com!newsfeed.nyu.edu!news.idt.net!feed2.news.erols.com!erols!woodstock.news.demon.net!demon!news.demon.co.uk!demon!mail2news.demon.co.uk!tnglwood.demon.co.uk!unclebob From: Robert Billing Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sun, 19 Apr 98 10:20:07 GMT Message-ID: <892981207snz@tnglwood.demon.co.uk> References: <3533F2F7.59E2A36D@ccw.ch> <6ha5h2$9q$1@irk.zetnet.co.uk> <3539420E.C57D3DD2@ccw.ch> Reply-To: unclebob@tnglwood.demon.co.uk X-Mail2News-User: unclebob@tnglwood.demon.co.uk X-Mail2News-Path: post-20.mail.demon.net!post.mail.demon.net!tnglwood.demon.co.uk X-Trace: mail2news.demon.co.uk 893070621 3142 unclebob tnglwood.demon.co.uk X-Complaints-To: abuse@demon.net X-Newsreader: Demon Internet Simple News v1.29 Lines: 28 In article <3539420E.C57D3DD2@ccw.ch> Neil.Franklin.remove.this@ccw.ch "Neil Franklin" writes: > Actually I can offer annother point: the pdp11 had demand paging (8k > pages IIRC), the 8086 doesn't, so that is what made it VM capable. Correct. The virtual address space is 8 pages of 8k, replicated for separate instruction and data spaces, and for different instruction mappings, on the top end models. The active part of each page can be anything between 64 bytes and the full 8k, starting at either end of the 8k block. Each page can also be read only, read-write, or nonresident, and read write pages can be marked dirty by any write. Pages used for heap space start at the bottom of the block, and are expanded upwards, pages used for stack are normally expanded downwards. Quite an impressive achievement for an architecture nearly 30 years old. -- I am Robert Billing, Christian, inventor, traveller, cook and animal lover, I live near 0:46W 51:22N. http://www.tnglwood.demon.co.uk/ "Bother," said Pooh, "Eeyore, ready two photon torpedoes and lock phasers on the Heffalump, Piglet, meet me in transporter room three" ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-dc.gip.net!news-peer.gip.net!news.gsl.net!gip.net!news-peer.sprintlink.net!news-peer-east.sprintlink.net!news.sprintlink.net!Sprint!rill.news.pipex.net!pipex!news.clara.net!peer.news.zetnet.net!zetnet.co.uk!not-for-mail From: lisard@zetnet.co.uk Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 19 Apr 1998 16:34:00 GMT Lines: 39 Message-ID: <6hd91o$93i$1@irk.zetnet.co.uk> References: <3539420E.C57D3DD2@ccw.ch> NNTP-Posting-Host: man-134.dialup.zetnet.co.uk X-Everything: Net-Tamer V 1.08X On 1998-04-19 Neil.Franklin.remove.this@ccw.ch said: :> what about the situation where you're using address remapping to :>pretend there's less memory than there actually is? (eg. to fit a :>22-bit memory space into a 16-bit virtual address space) all the :>dec literature called this virtual memory, didn't it? :Eaarrmmmm. You got me on that one :-) tee hee :> :I am going to try this escape: On an pdp11 you have multiple :processes of each 16 bit address range. If you consider the process :ID part of an "extended address" you still have mapping downwards. :That is, assuming enough processes. Else you would have not thrown :up the money for the RAM to use 22 bit addresses. With more RAM than :address space being used you essentially would be not using VM, :even if the architecture did support its use. yes, but you'd still always be stuck with a 16-bit virtual address space. however, paging is probably nicer to program with than segmentation as propounded by the 8086 (although whether it would be nicer than base+limit mapped into a flat space is a different matter - anyone care to comment?) :Without this escape an 8086 with its 16->20 bit stuff would be VM. :And I definitely do not want to give it that status :-) well, no, because the 8086 was 16+16->20. it's also generally bloody horrible :> :Actually I can offer another point: the pdp11 had demand paging (8k :pages IIRC), the 8086 doesn't, so that is what made it VM capable. yes, that we'd agree with quite happily. -- Communa (together) we remember... we'll see you falling you know soft spoken changes nothing to sing within her... ###### Newsgroups: alt.folklore.computers From: "John D. Burleson" Subject: Re: OS with no distinction between RAM and HD ? X-Nntp-Posting-Host: help.stl.mo.boeing.com Content-Type: text/plain; charset=us-ascii Message-ID: <353F5A2A.33575385@boeing.com> Sender: nntp@news.boeing.com (Boeing NNTP News Access) Content-Transfer-Encoding: 7bit Organization: The Boeing Company - McDonnell Douglas - Huntsville Division References: <3533F2F7.59E2A36D@ccw.ch> <6ha5h2$9q$1@irk.zetnet.co.uk> <3539420E.C57D3DD2@ccw.ch> Mime-Version: 1.0 Date: Thu, 23 Apr 1998 15:11:38 GMT X-Mailer: Mozilla 4.05 [en] (X11; U; SunOS 5.5.1 sun4u) Lines: 57 Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newsgate.cistron.nl!het.net!newsfeed.uk.ibm.net!sackheads.org!ibm.net!news.freedom2surf.net!btnet-peer!btnet!newsfeed.direct.ca!Supernews73!Supernews60!supernews.com!uunet!in1.uu.net!xyzzy!nntp Neil Franklin wrote: > > lisard@zetnet.co.uk wrote: > > > > On 1998-04-15 Neil.Franklin.remove.this@ccw.ch said: > > :> Any (hardware) system that makes it look as if you have more RAM > > :> than there really is, provides VM. > > > > :Yes. Actually to be precise: makes it that you have more valid > > :address range than RAM address range. Malloc() allocates address > > :range, not RAM space on VM systems. > > > > what about the situation where you're using address remapping to pretend > > there's less memory than there actually is? (eg. to fit a 22-bit memory > > space into a 16-bit virtual address space) all the dec literature called > > this virtual memory, didn't it? > > Eaarrmmmm. You got me on that one :-) > Any system that allows individual processes to use address ranges that do not necessarily coincide with hardware memory addresses is providing virtual memory. The most common case these days is larger-address-space-than-RAM, but any system that provides on-the-fly address mapping is providing VM, even if it's mapping multiple processes with small address space into a larger memory. This was a big deal once-upon-a-time as it allowed for processes (and therefore programmmers) to assume that the process address space began at a fixed location (quite often zero) when writing the code, allowing multiple programs to reside in RAM at once, which greatly simplified context switching, etc. My favorite quick and dirty definitions of the terms "actual," "transparent" and "virtual" as applied to Computer Science are these (I would attribute my source, if I remembered...): Actual - You can see it, and it's there. Transparent - It's there, but you can't see it. Virtual - You can see it, but it isn't there. In its most general sense, the term "virtual memory" applies to any scheme which allows the addresses used by processes to differ from the actual hardware addresses holding the values. This term has been applied in the literature in almost every way imaginable, which is why my first question after being told a system has VM is "what kind?". I'm probably dating myself severely. I would provide quotes/references but my entire personal technical library is in boxes awaiting an office move... john -- John Burleson (mailto:john.d.burleson@boeing.com) Principal Software Engineer The Boeing Company/McDonnell Douglas Aerospace/Huntsville (205)922-7589 FAX:(205)922-4890 ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!newspump.monmouth.com!newspeer.monmouth.com!newshub2.home.com!news.home.com!cpk-news-hub1.bbnplanet.com!cam-news-hub1.bbnplanet.com!news.bbnplanet.com!newsfeed.xcom.net!news.ultranet.com!not-for-mail From: "dave porter" Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 23 Apr 1998 17:12:03 GMT Organization: none of the above Lines: 74 Message-ID: <01bd6eda$d9f30430$0ba17392@glastonbury> References: <3533F2F7.59E2A36D@ccw.ch> <6ha5h2$9q$1@irk.zetnet.co.uk> <3539420E.C57D3DD2@ccw.ch> <353F5A2A.33575385@boeing.com> NNTP-Posting-Host: 146.115.161.11 X-Complaints-To: abuse@ultra.net X-Ultra-Time: 23 Apr 1998 17:12:03 GMT X-Newsreader: Microsoft Internet News 4.70.1161 I think you're confusing "virtual addressing" with "virtual memory". At least, the way I always learned the terms (back in the 1970s) was that any old machine with at least a base-and-limit register was providing virtual addressing (i.e., access to real memory through virtual or logical addresses) to its programs, but virtual memory specifically referred to techniques for allowing pieces of an address space to be moved to slower storage transparently to the program (the "transparently" part distinguishes this from overlay schemes, etc). dave -- For email, please remove the 'w' from my address. Sorry. John D. Burleson wrote in article <353F5A2A.33575385@boeing.com>... > Neil Franklin wrote: > > > > lisard@zetnet.co.uk wrote: > > > > > > On 1998-04-15 Neil.Franklin.remove.this@ccw.ch said: > > > :> Any (hardware) system that makes it look as if you have more RAM > > > :> than there really is, provides VM. > > > > > > :Yes. Actually to be precise: makes it that you have more valid > > > :address range than RAM address range. Malloc() allocates address > > > :range, not RAM space on VM systems. > > > > > > what about the situation where you're using address remapping to pretend > > > there's less memory than there actually is? (eg. to fit a 22-bit memory > > > space into a 16-bit virtual address space) all the dec literature called > > > this virtual memory, didn't it? > > > > Eaarrmmmm. You got me on that one :-) > > > > Any system that allows individual processes to use address ranges that > do not necessarily coincide with hardware memory addresses is providing > virtual memory. > The most common case these days is larger-address-space-than-RAM, but > any system that provides on-the-fly address mapping is providing VM, > even if it's mapping multiple processes with small address space into a > larger memory. This was a big deal once-upon-a-time as it allowed for > processes (and therefore programmmers) to assume that the process > address space began at a fixed location (quite often zero) when writing > the code, allowing multiple programs to reside in RAM at once, which > greatly simplified context switching, etc. > > My favorite quick and dirty definitions of the terms "actual," > "transparent" and "virtual" as applied to Computer Science are these (I > would attribute my source, if I remembered...): > > Actual - You can see it, and it's there. > Transparent - It's there, but you can't see it. > Virtual - You can see it, but it isn't there. > > In its most general sense, the term "virtual memory" applies to any > scheme which allows the addresses used by processes to differ from the > actual hardware addresses holding the values. This term has been > applied in the literature in almost every way imaginable, which is why > my first question after being told a system has VM is "what kind?". I'm > probably dating myself severely. > > I would provide quotes/references but my entire personal technical > library is in boxes awaiting an office move... > > john ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!News.Amsterdam.UnisourceCS!masternews.telia.net!news-stkh.gip.net!news-lond.gip.net!news.gsl.net!gip.net!btnet-peer!btnet!rill.news.pipex.net!pipex!bore.news.pipex.net!pipex!grot.news.pipex.net!join.news.pipex.net!pipex!krypton.inbe.net!INbe.net!not-for-mail From: lucvdv@null.net (Luc Van der Veken) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Sun, 10 May 1998 10:50:52 GMT Organization: . Lines: 53 Message-ID: <35568479.3057816@news.innet.be> References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> <6gpqgp$468$1@eldritch.dyn.ml.org> <3536410A.9A6D163B@boeing.com> NNTP-Posting-Host: pool02b-194-7-226-38.uunet.be Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/32.451 X-No-Archive: yes "John D. Burleson" told us > Ben Hutchings wrote: > > > > In article <6gou4r$s5q@chile.earthlink.net>, > > Joel C. Ewing wrote: > > >Adam Stouffer wrote: > > >The usual connotation of "swap file" is that all memory associated with > > >a suspended task is moved to auxiliary storage, and for the task to > > >resume execution, all its storage must first be restored to real > > >memory. > > Not to me. I think it can also be a file to which pages of memory are > > swapped. This is certainly the terminology that Linux uses. > > Properly speaking, this is paging and swapping means moving the whole > process as described. As various Unices began to add paged VM systems , > the swap partition was used as paging space. Therefore the name was > retained for all the wrong reasons and on modern UNIX variants (since > the early 80's, I would guess) most folks seem to use the terms > interchangeably. A couple of weeks too late, but I just bounced into this definition (confirming what you said) in the Linux system administrator's guide (sag-0.6-html/node60.html, Lars Wizenius, Nov '97): "A note on operating system terminology: computer science usually distinguishes between swapping (writing the whole process out to swap space) and paging (writing only fixed size parts, usually a few kilobytes, at a time). Paging is usually more efficient, and that's what Linux does, but traditional Linux terminology talks about swapping anyway." He also defines _virtual memory_ as "using a disk as an extension of RAM so that the effective size of usable memory grows correspondingly" Quite a difference from calling address remapping virtual memory (as was done a couple of posts back in this thread). A little side note touching the original subject: in Linux (and most unices, I guess) there is a file called /proc/kmem. This is a file that doesn't exist on any disk: instead it is an image of physical RAM, so you could say there's no distinction between RAM and HD (but not exactly in the same sense the original poster meant it: that would require it to be a directory instead of a file). I seem to remember that Coherent had both 'kmem' and 'mem': iirc, one allowed you to access memory in 1k blocks, the other allowed you to access a single byte. ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!ubnnews.unisource.ch!News.Amsterdam.UnisourceCS!newshunter.cosy.sbg.ac.at!newspump.monmouth.com!newspeer.monmouth.com!howland.erols.net!Supernews73!supernews.com!Supernews69!not-for-mail From: genew@vip.net (Gene Wirchenko) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Mon, 11 May 1998 06:24:26 GMT Organization: All USENET -- http://www.Supernews.com Lines: 26 Message-ID: <35563f04.1342224@news.vip.net> References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> <6gpqgp$468$1@eldritch.dyn.ml.org> <3536410A.9A6D163B@boeing.com> <35568479.3057816@news.innet.be> Reply-To: genew@vip.net NNTP-Posting-Host: 204.209.212.36 X-Trace: 894867763 A01OARAUVD424CCD1C usenet53.supernews.com X-Complaints-To: newsabuse@supernews.com X-Newsreader: Forte Free Agent 1.1/32.230 lucvdv@null.net (Luc Van der Veken) wrote: [snip] >A little side note touching the original subject: in Linux (and >most unices, I guess) there is a file called /proc/kmem. >This is a file that doesn't exist on any disk: instead it is an >image of physical RAM, so you could say there's no distinction >between RAM and HD (but not exactly in the same sense the >original poster meant it: that would require it to be a directory >instead of a file). >I seem to remember that Coherent had both 'kmem' and 'mem': iirc, >one allowed you to access memory in 1k blocks, the other allowed >you to access a single byte. I ran across mem and kmem in SVR2. IIRC, it was user memory and all memory including the kernel. mem was ordinary user accessible; kmem required root permissions. Sincerely, Gene Wirchenko C Pronunciation Guide: y=x++; "wye equals ex plus plus semicolon" x=x++; "ex equals ex doublecross semicolon" ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news.gsl.net!gip.net!newsfeed.internetmci.com!164.67.42.145!awabi.library.ucla.edu!132.239.1.220!ihnp4.ucsd.edu!news1.ucsd.edu!karoshi.ucsd.edu!not-for-mail From: brian@karoshi.ucsd.edu (Brian Kantor) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: 11 May 1998 08:44:23 -0700 Organization: The Avant-Garde of the Now, Ltd. Lines: 17 Approved: fi Message-ID: <6j76cn$hoo@karoshi.ucsd.edu> References: <3536410A.9A6D163B@boeing.com> <35568479.3057816@news.innet.be> <35563f04.1342224@news.vip.net> NNTP-Posting-Host: karoshi.ucsd.edu The "Pick" operating system, circa 1975 or so, indexed everything as disk-frame/offset. From the programmer's point of view, there was no RAM, even when you were writing in assembly. At that time it was sold as a "Microdata REALITY" system --- a Microdata 820 with a special firmware set; later it appeared in Honeywell hardware and further along it migrated to PC hardware, Unix, and elsewhere. Apparently it's still going strong in some niches of the business world. The 820 typically had 8-32K of core - real magnetic core - and the REALITY OS took advantage of that - for shows, we'd just halt the processor, power everything off, put it in a truck and after it was installed on the show floor, power it up and it would resume where it had left off. I left the Pick world in about 1979 but it took me until last year to get off the various mailing lists for used hardware. - Brian ###### Path: ccw.ch!elna.ethz.ch!news-zh.switch.ch!news.belnet.be!news-penn.gip.net!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!dispose.news.demon.net!demon!rill.news.pipex.net!pipex!bore.news.pipex.net!pipex!join.news.pipex.net!pipex!krypton.inbe.net!INbe.net!not-for-mail From: lucvdv@null.net (Luc Van der Veken) Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Mon, 11 May 1998 20:09:44 GMT Organization: . Message-ID: <35584f5f.825106@news.innet.be> References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> <6gpqgp$468$1@eldritch.dyn.ml.org> <3536410A.9A6D163B@boeing.com> <35568479.3057816@news.innet.be> <35563f04.1342224@news.vip.net> NNTP-Posting-Host: pool02b-194-7-226-55.uunet.be Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/32.451 X-No-Archive: yes Lines: 33 genew@vip.net (Gene Wirchenko) told us > lucvdv@null.net (Luc Van der Veken) wrote: > [snip] > >I seem to remember that Coherent had both 'kmem' and 'mem': iirc, > >one allowed you to access memory in 1k blocks, the other allowed > >you to access a single byte. > > I ran across mem and kmem in SVR2. IIRC, it was user memory and > all memory including the kernel. mem was ordinary user accessible; > kmem required root permissions. Possible - it's been too long (I must still have the Coherent manual somewhere, just don't remember where). Otoh, I seem (again not 100% sure) to remember a note explaining why the two forms were there: kmem allowed you to access all memory the easy way, while mem required long record numbers (in the first version(s) of Coherent things were 16 bit by default: 64k limits on all segments, even though it used all of your computer's memory, also above 1MB). I suppose a user-accessible 'mem' would exclude other users memory also, not just the kernel. I'm not saying it wasn't so, but it seems difficult to implement that way (thinking of fragmented memory). And then, maybe both ways existed and things just differed between SVR2 and Coherent (which looked like unix, but wasn't it). Somebody around with a better memory than me? ###### Path: ccw.ch!usenet From: Neil Franklin Newsgroups: alt.folklore.computers Subject: Re: OS with no distinction between RAM and HD ? Date: Tue, 12 May 1998 00:52:09 +0200 Organization: My own Private Self Lines: 25 Message-ID: <35578119.5755FF71@ccw.ch.remove> References: <6gnu0q$np8$1@max82.public.ox.ac.uk> <352FD629.7565@NO-SPAM.sgi.net> <6gou4r$s5q@chile.earthlink.net> <6gpqgp$468$1@eldritch.dyn.ml.org> <3536410A.9A6D163B@boeing.com> <35568479.3057816@news.innet.be> <35563f04.1342224@news.vip.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.05 [en] (X11; I; Linux 2.0.27 i486) Gene Wirchenko wrote: > > lucvdv@null.net (Luc Van der Veken) wrote: > > >A little side note touching the original subject: in Linux (and > >most unices, I guess) there is a file called /proc/kmem. > > I ran across mem and kmem in SVR2. IIRC, it was user memory and > all memory including the kernel. mem was ordinary user accessible; > kmem required root permissions. Actually /dev/kmem is defined as an character device that scans through the physical RAM. Usually used by "ps" and similar in the days before /proc appeared. Today that is used for "ps". OTOH there now is /proc/kcore that shows physical RAM as an pseudo file. /dev/men is an character device that scans the virtual address space of the active process, that is the process opening it. Usually used by debuggers to manipulate the process they are controlling. Also superseeded by /proc//mem. -- private: Neil.Franklin@ccw.ch.remove http://www.ccw.ch/Neil.Franklin/ office: franklin@arch.ethz.ch.remove http://caad.arch.ethz.ch/~franklin/ Lawyers are killing society, perhaps we should return the favour.