peterk@cbmger.UUCP (Peter Kittel GERMANY) (03/25/91)
Cited from an ad about harddisk controllers in the German magazine "Kickstart" (April issue, backpage): In the course of a severe software revision, EVOLUTION sets another milestone of power. By using the MMU, as present on most 68020 and all 68030 cards, you may use as much harddisk storage as you want as FAST RAM. The extremely high data transfer rate of EVOLUTION (beyond 2 MB/s) and most current page exchange algorithms (similar to UNIX) open new possibilities of use. Company: Macro System, Witten, Germany, tel. 0049-2302-27073 Now this is April's issue, but this is a commercial ad. I didn't talk with them, but this sounds hot. I also don't know whether they developed this themselves or only distribute it. It may even be an imported product? -- Best regards, Dr. Peter Kittel // E-Mail to \\ Only my personal opinions... Commodore Frankfurt, Germany \X/ {uunet|pyramid|rutgers}!cbmvax!cbmger!peterk
GUTEST8@cc1.kuleuven.ac.be (Ives Aerts) (03/27/91)
I don't think it's an April joke. Amiga Magazin allso had a small article about it. I don't know the details of it but it said something about being able to use virtual memory if you have their harddisk controller and an MMU. ------------------------------------------------------------------------ Ives Aerts | IBM definition SY-34378 GUTEST8@BLEKUL11.BITNET | A signature consists of sequences of gutest8@cc1.kuleuven.ac.be | non-blank characters separated by blanks. ------------------------------------------------------------------------
stefanb@kaa.informatik.rwth-aachen.de (Stefan Becker) (03/27/91)
peterk@cbmger.UUCP (Peter Kittel GERMANY) writes: >as FAST RAM. The extremely high data transfer rate of EVOLUTION >(beyond 2 MB/s) and most current page exchange algorithms (similar >to UNIX) open new possibilities of use. Virtual RAM? Paging? That does remind me of the 4th grade project of Mr. Valentin Pop. (Sorry can't remember your name :-) from C=. What was his experience? 98% of the Amiga programs didn't run, because they expect REAL memory. Facit: You gain NOTHING. Nice marketing idea, though :-) Stefan BTW: Peter, do you really expect a WORKING Virtual RAM system from a german firm like MacroSystems? They usually employ 64'er hardware freaks and SEKA programmers. -- Mail : Stefan Becker, Holsteinstrasse 9, D-5100 Aachen /// Only Phone : +49-241-505705 FIDO: 2:242/7.6 Germany /// Amiga makes Domain: stefanb@informatik.rwth-aachen.de \\\/// it possible.. Bang : ..mcvax!unido!rwthinf!stefanb \XX/ -->A3000/25<--
kris@tpki.toppoint.de (Kristian Koehntopp) (03/28/91)
peterk@cbmger.UUCP (Peter Kittel GERMANY) writes: [ ... Availability of VM for Amiga ... ] >Now this is April's issue, but this is a commercial ad. I didn't >talk with them, but this sounds hot. I also don't know whether they >developed this themselves or only distribute it. It may even be an >imported product? This is no april's joke at all, but instead a program smaller than 10 KB plus two 250 byte utility programs. I have no idea, wether these are depending on the evolution hardware. I have been able to test the Evolution HDD Controller on an Amiga 2500 with 2 MB Fast Memory and Quantum 50 MB SCSI HDD. I was using the entire disk as VM and tested several applications with this amount of memory. Most memory-clocks fail, since they used signed shorts for displaying kilobytes of memory and having an entire 50 MB HDD as VM gives you negative numbers on these displays. Myclock had to be slightly patched and recompiled for proper operation. Installation was easy and straightforward. I chose the entire disk for VM and the entire 2 MB as buffer pool for the VM. There is no use in having pure, no-VM fast memory, as long as you don't have any realtime applications. These better go in real RAM and be not paged out, to achieve good and somewhat determined response times. DPaint III worked good with this immense amount of mem, although allocating memory for a large animated brush lasted nearly one minute. I suppose this is because DPaint III also cleares this memory. SnapShot Studio plus, Version 5, ran fine and was faster at allocating memory for a 50 MB RAM animation. Of course the overall playback speed was nearly the same as for a disk animation, but when it came to browsing forward and backward through a small set of frames, response was real fast. Several other applications also worked fine except for some minor problems with short integers for free memory displays. These problems will not show up, if you choose a sane amount of VM, instead going for manta-style amounts ("Ey, my A2500 now displays 500 MB of free ram at boot and I already have 3 resident copies of CED running then."). Yours, Kristian PS: Commo should really include this in the next KS release. Since the program is that small, it should not very difficult to handle. Kristian Koehntopp, Harmsstrasse 98, 2300 Kiel, +49 431 676689 "Im uebrigen ist 'Z-NETZ-Sysops quaelen' auf Dauer weder lustig noch befriedigend." - sysop@infinet.zer.sub.org
sbeagle@kennels.actrix.gen.nz (Sleeping Beagle) (03/30/91)
stefanb@kaa.informatik.rwth-aachen.de (Stefan Becker) writes: > peterk@cbmger.UUCP (Peter Kittel GERMANY) writes: > >as FAST RAM. The extremely high data transfer rate of EVOLUTION > >(beyond 2 MB/s) and most current page exchange algorithms (similar > >to UNIX) open new possibilities of use. > > Virtual RAM? Paging? That does remind me of the 4th grade project of > Mr. Valentin Pop. (Sorry can't remember your name :-) from C=. > > What was his experience? 98% of the Amiga programs didn't run, because > they expect REAL memory. > > Facit: You gain NOTHING. Nice marketing idea, though :-) Why shouldn't virtual RAM work on the Amiga? It works on the PC and Macintosh.... Is there any real reason, or is it just that no one got around to working out the bugs... -- Sleeping Beagle (aka Thomas Farmer) sbeagle@kennels.actrix.gen.nz The Kennels Ph. +64-4-796306 (voice) 25 Awarua St, Ngaio, Wellington, New Zealand. "You ain't nothin' but a Hound Dog."
holgerl@amiux.UUCP (Holger Lubitz) (04/01/91)
In article <wDskZ1w164w@kennels.actrix.gen.nz> sbeagle@kennels.actrix.gen.nz (Sleeping Beagle) writes: >Why shouldn't virtual RAM work on the Amiga? It works on the PC >and Macintosh.... Neither the PC nor the Macintosh multitask like the Amiga does. Banking the virtual memory works just fine on a PC, since there is only one program running in the machine, and the processor has no need to wildly jump back and forth while switching between tasks. But did you ever see Windows 3.0 on a 386 with not so much memory (the one I used had "only" 2 MB) trying to get two programs running ? The harddisk runs and runs and runs and runs all the time, and system performance degrades noticeably. Just because the system has to load from the harddisk most of the time. And even Windows is not the type of multitasking the Amiga has. No nifty messages flowing through your system, no inter-task communication (as far as I heard from a PC programmer). Merely some kind of very comfortable task switching. And regarding the Mac: Virtual Memory is *announced*, but System 7 is not released yet, and regarding what I have heard from beta testers, it is not as nicely done as they expected it to be. Virtual Memory is nice with fast harddisks, and enough REAL memory to run your programs in. But remember: You always have to GIVE UP some of your real memory to allow virtual memory. You cannot access some sector on your harddisk directly with the processor. You can only use some clever software using the MMU to determine if the processor tries to access a segment of virtual memory, then to check if the correct contents are already in your reserved segment of real memory, to copy the right contents from the harddisk into the reserved segment if they are not, then redirecting the access there. Tricky. And if you ever run out of real memory so that the OS starts placing YOUR PROGRAMS in virtual memory, be prepared for a severe performance loss. You still need some clever caching techniques to prevent your software to reload data areas every now and then if two different programs run in multitasking keep accessing two different data areas in virtual memory, far enough from each other that each one resides in different segments. Virtual memory is good for keeping megabytes of data, ok. But then you could also keep them in a file. Of course it would be nice to have VM in the OS, but don't expect that you can simply add your harddisks capacity to your main memory with it. Best regards, Holger -- Holger Lubitz | holgerl@amiux.uucp Kl. Drakenburger Str. 24 | holgerl@amiux.han.de D-W-3070 Nienburg | cbmvax.commodore.com!cbmehq!cbmger!amiux!holgerl
gilgalad@caen.engin.umich.edu (Ralph Seguin) (04/01/91)
In article <wDskZ1w164w@kennels.actrix.gen.nz> sbeagle@kennels.actrix.gen.nz (Sleeping Beagle) writes: >stefanb@kaa.informatik.rwth-aachen.de (Stefan Becker) writes: > >> peterk@cbmger.UUCP (Peter Kittel GERMANY) writes: >> >as FAST RAM. The extremely high data transfer rate of EVOLUTION >> >(beyond 2 MB/s) and most current page exchange algorithms (similar >> >to UNIX) open new possibilities of use. >> >> Virtual RAM? Paging? That does remind me of the 4th grade project of >> Mr. Valentin Pop. (Sorry can't remember your name :-) from C=. >> What was his experience? 98% of the Amiga programs didn't run, because >> they expect REAL memory. What are you talking about? You are accessing real memory, it's just mapped into a virtual address space. All address translations are handled by the MMU. >> Facit: You gain NOTHING. Nice marketing idea, though :-) What? There is a hell of a lot to be gained with virtual memory. Ever hear of a debugging environment where the system doesn't lock itself up when a process does? Segmentation fault... :) >Why shouldn't virtual RAM work on the Amiga? It works on the PC >and Macintosh.... Hmmm... you mean with some sort of virtual memory software of course :) What's the point of having virtual memory under DOS anyways :) ? Ralph Seguin gilgalad@caen.engin.umich.edu 536 South Forest Apt. #915 gilgalad@zip.eecs.umich.edu Ann Arbor, MI 48104 (313) 662-4805
dege@cs.umn.edu (Dege Jeffrey Charles) (04/01/91)
In <holgerl.0613@amiux.UUCP> holgerl@amiux.UUCP (Holger Lubitz) writes: >In article <wDskZ1w164w@kennels.actrix.gen.nz> sbeagle@kennels.actrix.gen.nz (Sleeping Beagle) writes: >>Why shouldn't virtual RAM work on the Amiga? It works on the PC >>and Macintosh.... >Neither the PC nor the Macintosh multitask like the Amiga does. >Banking the virtual memory works just fine on a PC, since there is only >one program running in the machine, and the processor has no need to wildly >jump back and forth while switching between tasks. >But did you ever see Windows 3.0 on a 386 with not so much memory (the >one I used had "only" 2 MB) trying to get two programs running ? The harddisk >runs and runs and runs and runs all the time, and system performance degrades >noticeably. In a simple test of Windows 3.0 on a 2 Meg 386sx, a make under Windows took 1Hr and 38 minutes, while the same make under DOS took 4 minutes and change. Most of this seemd to be due to heavy disk swapping. -----------------------
mwm@pa.dec.com (Mike (My Watch Has Windows) Meyer) (04/02/91)
In article <holgerl.0613@amiux.UUCP> holgerl@amiux.UUCP (Holger Lubitz) writes:
Tricky. And if you ever run out of real memory so that the OS starts placing
YOUR PROGRAMS in virtual memory, be prepared for a severe performance loss.
As opposed to what happens when you run out of real memory with no VM.
If you're real lucky, you've got a well-written application that will
pause while you find out if you've got enough memory get to a window
and run a delete to make disk space to save things into. More likely,
you get to start finding bugs in your application, and the OS.
Paging/swapping is a "fail-soft" option. It takes you from not being
able to perform a task, to being able to perform a task at a reduced
rate of speed (when compared to useing real memory). The cost is a
degredation in speed right at the edge of failure (*), and that's it.
VM is a tool. Like any other tool, VM used wisely can save you time
and/or money. Used badly, it can make your life miserable. I know
I've got a number of tasks that don't fit on my system.
<mike
(*) It's demonstrably possible to design an MMU that doesn't eat clock
cycles on a machine designed to work with an MMU. There are those who
claim that designing a machine to use an MMU makes it slower. The
truth of this is unknown, and difficult to test.
--
Lather was thirty years old today, Mike Meyer
They took away all of his toys. mwm@pa.dec.com
His mother sent newspaper clippings to him, decwrl!mwm
About his old friends who'd stopped being boys.
jdickson@jato.jpl.nasa.gov (Jeff Dickson) (04/02/91)
>Why shouldn't virtual RAM work on the Amiga? It works on the PC >and Macintosh.... > >Is there any real reason, or is it just that no one got around >to working out the bugs... > >-- > Sleeping Beagle (aka Thomas Farmer) sbeagle@kennels.actrix.gen.nz > The Kennels Ph. +64-4-796306 (voice) > 25 Awarua St, Ngaio, Wellington, New Zealand. > "You ain't nothin' but a Hound Dog." Virtual memory is not possible on any Amiga that does not have an MMU. Without one, there is no way to "catch" references to memory that is swapped out. -jeff
peterk@cbmger.UUCP (Peter Kittel GERMANY) (04/02/91)
In article <1991Apr1.222906.12714@jato.jpl.nasa.gov> jdickson@jato.Jpl.Nasa.Gov (Jeff Dickson) writes: > Virtual memory is not possible on any Amiga that does not have an >MMU. Without one, there is no way to "catch" references to memory that is >swapped out. Yes. And the said software from EVOLUTION says that it *needs* a MMU. So this does run on *the* Amiga, but not on *any* Amiga. -- Best regards, Dr. Peter Kittel // E-Mail to \\ Only my personal opinions... Commodore Frankfurt, Germany \X/ {uunet|pyramid|rutgers}!cbmvax!cbmger!peterk
walrus@wam.umd.edu (Udo K Schuermann) (04/03/91)
In article <1032@cbmger.UUCP> peterk@cbmger.UUCP (Peter Kittel GERMANY) writes: > Yes. And the said software from EVOLUTION says that it *needs* a MMU. > So this does run on *the* Amiga, but not on *any* Amiga. Well, I've got an MMU, and I want virtual memory! My questions: a) when is EVOLUTION available? b) how much will it cost? c) will it live with my GVP SCSI controller? d) will it operate under 2.0 when it becomes available for the 2000? Thanks for any info! ._. Udo Schuermann And now for something completely different: ( ) walrus@wam.umd.edu An Iraqi dictator without his underpants.
peterk@cbmger.UUCP (Peter Kittel GERMANY) (04/03/91)
In article <1991Apr3.053350.20366@wam.umd.edu> walrus@wam.umd.edu (Udo K Schuermann) writes: >In article <1032@cbmger.UUCP> peterk@cbmger.UUCP (Peter Kittel GERMANY) writes: >> Yes. And the said software from EVOLUTION says that it *needs* a MMU. I should add something I heard: This only works, if the MMU is currently totally free, unused. If it is already in use (like in today's A3000), then the software can't provide VM. >Well, I've got an MMU, and I want virtual memory! >My questions: > a) when is EVOLUTION available? *** seems now > b) how much will it cost? *** controller without hd: 448,- DM; up to 5588,- DM with a 660 MB drive (today 1 US$ ca. equals 1,70 DM) > c) will it live with my GVP SCSI controller? *** it is an own controller, I fear you won't get the software separately > d) will it operate under 2.0 when it becomes available for the 2000? *** they say so contact: MacroSystem, Billerbeckstr. 39a, 5810 Witten, Germany Tel. 0049 - 2302 - 27073, Fax ... - 27072 (also offering Medusa, the ST emulator and much other hw stuff) -- Best regards, Dr. Peter Kittel // E-Mail to \\ Only my personal opinions... Commodore Frankfurt, Germany \X/ {uunet|pyramid|rutgers}!cbmvax!cbmger!peterk
rivero@dev8.mdcbbs.com (04/03/91)
In article <1032@cbmger.UUCP>, peterk@cbmger.UUCP (Peter Kittel GERMANY) writes: > In article <1991Apr1.222906.12714@jato.jpl.nasa.gov> jdickson@jato.Jpl.Nasa.Gov (Jeff Dickson) writes: >> Virtual memory is not possible on any Amiga that does not have an >>MMU. Without one, there is no way to "catch" references to memory that is >>swapped out. > > Yes. And the said software from EVOLUTION says that it *needs* a MMU. > So this does run on *the* Amiga, but not on *any* Amiga. > > -- > Best regards, Dr. Peter Kittel // E-Mail to \\ Only my personal opinions... > Commodore Frankfurt, Germany \X/ {uunet|pyramid|rutgers}!cbmvax!cbmger!peterk Even with an MMU, the way Amigados structures the free lists in its memory manager produces accumulating problems with a virtual memory manager, either in software or hardware. That the memory tends to fragment rapidly has always been one of the Amigas problems. Michael P.S. Flamers! I LOVE the Amiga. I Have 3 of them at home. Every time I make a critical comment about the Amiga I get a ton of bad mouth E-Mail I do not feel I deserve. Even a loving parent should be able to admit his kid has pimples!
rbabel@babylon.rmt.sub.org (Ralph Babel) (04/04/91)
In article <3033@tpki.toppoint.de>, kris@tpki.toppoint.de (Kristian Koehntopp) writes: > I have no idea, wether these are depending on the > evolution hardware. They depend on driver-private entry points in the Evolution software (disk I/O without OS). > Several other applications also worked fine [...] Try it under some really heavy system load. Or what about SCSIF_AUTOSENSE to virtual memory? Why not backup the SCSI harddrive (the one used for paging) to a SCSI tape drive using a backup program allocating buffers in virtual memory? Then, while backing up the system, eject the tape cartridge. Ralph
rbabel@babylon.rmt.sub.org (Ralph Babel) (04/04/91)
In article <1991Apr3.053350.20366@wam.umd.edu>, walrus@wam.umd.edu (Udo K Schuermann) writes: > c) will it live with my GVP SCSI controller? Amiga-OS 1.3 does not really support DMA _and_ MMU at the same time. Ever tried to save a SetCPU KICKROM image? :-) Ralph
kris@tpki.toppoint.de (Kristian Koehntopp) (04/05/91)
walrus@wam.umd.edu (Udo K Schuermann) writes: >Well, I've got an MMU, and I want virtual memory! >My questions: > a) when is EVOLUTION available? > b) how much will it cost? > c) will it live with my GVP SCSI controller? > d) will it operate under 2.0 when it becomes available for the 2000? I have been mailed by so many people, so I am going to post complete information on this device here. The Evolution HDD Controller comes with a workbench disk including the following additional programs: Directory "EvolutionInstall 2.1:" on Wednesday 03-Apr-91 MedusaSCSI 2260 ----rwed 29-Jul-90 21:11:54 -> Device for the "Medusa" Atari Emulator SCSIStop 1300 ----rwed 29-Jan-91 02:12:37 -> Stops all HD motors SCSIStop.info 1208 ----rwed 26-Feb-91 02:04:08 SCSIInstall 10040 ----rwed 07-Feb-78 01:53:04 -> prep & format SCSIInstall.info 1118 ----rwed 26-Feb-91 02:04:03 SCSIBootSel 4120 ----rwed 07-Feb-78 01:53:19 -> Select Boot HD and Partition SCSIBootSel.info 1118 ----rwed 26-Feb-91 02:03:57 VMem Dir ----rwed 19-Mar-91 12:31:16 -> Directory with VM programs VMem.info 712 ----rw-d 19-Mar-91 12:27:53 20 files - 14 directories - 103 blocks used Directory "EvolutionInstall 2.1:devs" on Wednesday 03-Apr-91 evolution.amhd 2220 ----rwed 26-Feb-91 02:16:07 -> Device driver for AMAX II 8 files - 3 directories - 113 blocks used Directory "EvolutionInstall 2.1:VMem" on Wednesday 03-Apr-91 PreferVirtualMem 256 ----rwed 26-Feb-91 02:07:37 -> Set priority of VM to one higher than Fast Memory PreferVirtualMem.info 1898 ----rwed 26-Feb-91 02:07:47 PreferNormalMem 256 ----rwed 26-Feb-91 02:07:54 -> Set priority of VM to one lower than Chip Memory PreferNormalMem.info 1898 ----rwed 26-Feb-91 02:08:32 VRam 8272 ----rwed 26-Feb-91 02:07:24 -> Vram Tool, MMU required VRam.info 2210 ----rw-d 19-Mar-91 12:28:24 7 files - 40 blocks used TOTAL: 200 files - 24 directories - 1741 blocks used The Evolution HD Controller can be obtained from Macrosystems Billerbeckstrasse 39a D - 5810 Witten Germany Tel.: +49 2302 27073 Fax.: +49 2302 27072 Mo-Fr: 9.00 AM to 12.30 PM 14.00 PM to 18.00 PM This is Central European Time (GMT +0100) and CET DST (GMT +0200) I do not work for Macrosystems nor do I sell their products. PLEASE DO NOT MAIL ME, IF YOU WANT TO HAVE IT. Maximum performance on an A2500 (68020) as measured by DiskPerf (Fish 187): - Imprimis Wren Runner 7: 2,24 MB/s - Quantum Prodrive 210S : more than 1.00 MB/s The Evolution HDD Controller is a Card Controller, the disk is mounted directly onto the card. The construction seems somewhat mechanically instable. The controller supports Burstmode transmission into memory, whatever this means, it surely is fast. All partitions can be automounted. Any partition can be boot partition. Kickstart 1.2, 1.3 and 2.x are supported. Documentation is in german language. I have no idea, whether english docs can be obtained. Prices are (netto without tax): Filecard (solo) 392.- DM / 230.- $ Filecard + 52MB Quantum LPS 910.- DM / 535.- $ Filecard + 105MB Quantum LPS 1490.- DM / 876.- $ Filecard + 170MB Quantum 2095.- DM / 1232.- $ Controller + ext. Wren Runner 7 (660 MB) 4901.- DM / 2883.- $ Medusa Atari Emulator with Tos 349.- DM / 205.- $ Add post & package. Don't forget your tax and custom. US $ price is calculated and may vary heavily, according to US politics :-) Again: I am not connected to Macrosystems. I do not sell their products. Any, yes, an MMU is required. Without an MMU this is just another fast SCSI controller. I have no idea, whether their VRam-Tool works with any other controller. PLEASE SNAIL-MAIL OR PHONE THEM, NOT ME. Kristian Koehntopp, Harmsstrasse 98, 2300 Kiel, +49 431 676689 "Das Editier, das Editier, das lebt jetzt leider nicht mehr hier." -- aus "Un-Zen oder die Kunst des schlechten Kalauers".
jesup@cbmvax.commodore.com (Randell Jesup) (04/05/91)
In article <1991Apr3.114837.1@dev8.mdcbbs.com> rivero@dev8.mdcbbs.com writes: >Even with an MMU, the way Amigados structures the free lists in its memory >manager produces accumulating problems with a virtual memory manager, either >in software or hardware. That the memory tends to fragment rapidly has always >been one of the Amigas problems. The tricky part is programs that a) access memory under Forbid or worse yet disable (they may avoid this by only working with their own controller, and not using interrupts (or requiring their use) in the driver); b) programs that put critical code in VM (interrupt handlers, for example). The "proper" method requires programs specify that they want VM and are willing to abide by the rules for safe access. All this was hashed out quite a while ago in c.s.a.tech. -- Randell Jesup, Keeper of AmigaDos, Commodore Engineering. {uunet|rutgers}!cbmvax!jesup, jesup@cbmvax.commodore.com BIX: rjesup Disclaimer: Nothing I say is anything other than my personal opinion. Thus spake the Master Ninjei: "To program a million-line operating system is easy, to change a man's temperament is more difficult." (From "The Zen of Programming") ;-)