Info-IBMPC@WSMR-SIMTEL20.ARMY.MIL (02/14/89)
Info-IBMPC Digest Mon, 13 Feb 89 Volume 89 : Issue 21 Today's Editor: Gregory Hicks - Chinhae Korea <COMFLEACT@Taegu-EMH1.army.mil> Today's Topics: RPICICGE.Bitnet Listing DIRs of files Closing files left open by TSR UNIX on NEC V20; 3.5" drives; DiskManager PC RE: Memory parity interupt error RE: Memory parity interupt error on a Zenith AT (IBM clone) More on random number generators Hard Disk Info Wanted Hard Disk Interleave Important update to DSZ fixes bug I/O on an Epson Equity II Waveform Editors for PC's Lotus 123 behaviour problem millisecond timing on PC More 4107 questions Re: Data acquisition for IBM PS2 model 80 SIMTEL20 index and CRC list files now available archived sound for ibm vga pallete X windows (server) for PC ---------------------------------------------------------------------- Date: Mon, 06 Feb 89 12:41:26 EST From: "John S. Fisher" <FISHER%RPICICGE.BITNET@VM.ECS.RPI.EDU> Subject: RPICICGE.Bitnet Listing DIRs of files To get a directory list of files more than 30 days old, the user needs to put a number (>30) after the file name pattern. For example, /PDDIR PD:<msdos.asmutl>*.* 9999 Note that there is a space before the 9999. ------------------------------ Date: Fri, 3 Feb 89 18:11 GMT From: "Barry Redmond" <BREDMOND%dit.ie@CUNYVM.CUNY.EDU> Subject: Closing files left open by TSR Can anybody tell me if there is a way to close a file that has been accidentally left open by a TSR? The file is a configuration file only required when the TSR is loaded, but it doesn't close it when it is finished using it. The TSR is loaded from a network file server, so the open file causes difficulties with trying to manipulate links to the server. Perhaps it is possible to track down the FCB? All suggestions welcome. Thanks in advance, Barry Redmond BREDMOND@DIT.IE Dept of Electronics & Communications Dublin Institute of Technology Kevin St Dublin 8 Ireland ------------------------------ Date: Sun, 5 Feb 89 19:25 EDT From: <J_KAZURA@UNHH> Subject: UNIX on NEC V20; 3.5" drives; DiskManager PC Greetings All, I just read Info-IBMPC Digest V89 #17, and I wanted to respond to a few questions and comments. 1) Don Meredith had a question about a UNIX system for a PC with an NEC V20. The Computer Science deptartment here at the University of New Hampshire uses a program called MINIX. It barely runs on the AT&T PC6300's but runs fine on my clone with an NEC V20 (8Mhz). It does emulate UNIX just fine. Janna Stevens here could give you more help on the matter if anyone is interested. JL_STEVENS@UNHH. 2) James Massey commented that he had had problems with 3.5" drives. He said that there were problems going from high density to low density and so forth. I thought I'd just say that we have had no problems (as yet) with switching media and drives. We have IBM PS/2's all models. Using DOS 3.3 3) A note about the DiskManager PC. We have a public cluster with 15 IBM Mod 30/286 E21 (20meg) running DiskManager PC the programs is a pain in the bottom for advanced users, but works great. Contact Janna Stevens for actual experience with the program. <JL_STEVENS @ UNHH> Joe Kazura University of New Hampshire J_KAZURA@UNHH ------------------------------ Date: Sun, 5 Feb 89 20:19 EST From: "Speed kills..." <INS_AJDC@JHUVMS> Subject: RE: Memory parity interupt error > I'm working on a project in the physics department. I'm trying to debug >an error of the form "memory parity interrupt." The code in which the >error occurs is a data acquistion program. > Data is input from a photosensor which records the intensity of light. >The data is written into memory via a D.M.A. (Direct Memory Access) >operation. The first time the data is manipulated in the code produces >the "memory parity interrupt." Actually, the error doesn't occur all the >time, only some of the time but at the same point of the code. >Can you tell be any information about the nature of this error? Ie, when >is the memory parity tested? What tests the parity? Steve, A parity error is displayed when the computer realizes that the memory has an error. Specifically it holds one extra bit per byte that is set to either a 1 or 0 depending on what the other 8 are set to. This is used as a checksum. And when the check fails, the computer "crashes". Possible sources of error are, a faulty DMA chips (they are not setting the parity bit right), a faulty memory chip (it loses some data), or faulty power supply (the power to the chips is unstable and the data "fades"), also the way you write to memory, if you by-pass certain steps (I don't know the exact procedure) you could be skipping the parity (unlikely, otherwise you would get lots of errors), and lastly large amounts of data will eventualy produce an error (the laws of probability). As a solution you might want to have the hardware checked, or try to break down the amounts of data you send. There is a short term software "solution", PC Magazine ran an article that contained a TSR program that would keep the computer from "crashing" when it detected a parity error, but alas I don't remeber what issue. -- Speed (Justin Chandler) INS_AJDC@JHUVMS G49I1930@JHUVM ------------------------------ Date: Sun, 5 Feb 89 18:51:23 PST From: madler@Hamlet.Bitnet (Mark Adler) Subject: RE: Memory parity interupt error on a Zenith AT (IBM clone) In response to: >> Data is input from a photosensor which records the intensity of >> light. The data is written into memory via a D.M.A. (Direct >> Memory Access) operation. The first time the data is manipulated >> in the code produces the "memory parity interrupt." Actually, the >> error doesn't occur all the time, only some of the time but at the >> same point of the code. >> >> Can you tell be any information about the nature of this error? >> Ie, when is the memory parity tested? What tests the parity? >> >> Steve Zelazny (v055l9y3@ubvms.bitnet) Parity checking provides single-bit error detection. An extra parity bit is saved every time a byte is written, and checked every time a byte is read. This applies to 16 bit reads and writes as well---two parity bits are generated or checked, one for each byte. All of this is done in the memory hardware. If the parity is wrong, a non-maskable interrupt is generated and the BIOS will service this interrupt and report a parity error. This is usually considered an extremely fatal error, since the byte that was read might have been an instruction. A parity error on a read indicates that one of the nine bits went wrong either when the byte was written, when it was read, or sometime in between. If it was sometime in between, that means that something is wrong with one of the memory chips that the nine bits are stored in. If it happened during the read or write, it may indicate that there is something wrong with the timing, most likely not enough time was allowed for the read or the write. In your case, since you have been using the memory for other things and haven't had any problem with it before, the parity error is probably not occuring during the read, or the interval before the read. I would guess that the timing of the DMA write operation being generated by your data acqusition hardware is not compatible with the Zenith memory for some reason. Assuming that the DMA device was designed and tested with a real IBM, the incompatibility probably can be blamed on your Zenith. Of course, blaming Zenith doesn't solve anything. I would first see if your data acqusition hardware has any adjustments for its memory timing---perhaps the insertion of a wait state. If it does, then do this and if my guesses are correct, your problem should go away. If not, and you have an 8 MHz Zenith, try lowering the clock rate to 6 MHz. This should also make the problem go away. If you can't do that, then the only thing left that doesn't require a digital hardware designer is to put in faster memory chips. That is, go from 150 ns RAM's to 120 ns RAM's. Unfortunately, that can be pretty expensive nowadays. If your problem is marginal (which it sounds like it is), this should make the problem go away as well. If you are, or have access to, someone who can design digital hardware, then there is the possiblilty of modifying the timing of the DMA writes from your data acqusition board. This is extremely difficult though and is frought with peril. (A fix for one problem can sometimes cause other unforseen problems.) It may be worth checking to see if the manufacturer of the board has a fix like this. There is one more thing you can do, but it doesn't solve the problem: you can disable the parity errors. Port 61 hex controls the speaker and parity error actions. The following code will ignore parity errors that are generated either on the system board or on the bus (from an assertion of the -I/O CH CK line): in al,61h or al,00001100b out 61h,al To turn parity checking back on: in al,61h and al,11110011b out 61h,al This would be the equivalent of accepting noise in your data that consists of occasional bit flips. It may not be so bad, since the most likely bit to be wrong out of the nine is actually the parity bit, which isn't part of your data anyway. The reason for this is that it takes a tiny bit more time to generate the parity bit, so it gets to the RAM's a little later than the other eight data bits. Therefore, if there is not quite enough time to get that parity bit written, it will cause a parity error even though the other eight bits (the data) are correct. Mark Adler bitnet: madler@hamlet arpa: madler@hamlet.caltech.edu ------------------------------ Date: Mon, 6 Feb 89 13:12:52 PST From: madler@Hamlet.Bitnet (Mark Adler) Subject: More on random number generators This is a comment on John Levine's contribution on random number generators ... He quotes the article by Park and Miller in Communications of the ACM volume 31, number 10 (October 1988): "Random Number Generators: Good Ones Are Hard To Find". Quite true. Upon looking up the article, however, I was dissapointed to find no new results beyond what is in Knuth (Seminumerical Algoithms, 2nd ed.) The authors, who teach a computer simulation course, have found that the random number generators in many computer science textbooks are remarkably bad. They therefore propose a generator, with several desireable features, as a "standard" generator, at least for teaching purposes. Among those features are: randomness (of course), portability, and repeatability. This is to facilitate checking the results of students and independence from random number generators that come with particular languages or machines (which are also usually bad). The random number generator they suggest in the article, the one that John included in his contribution, has the great virtue of being portable across machines and languages that have 32 bit (or larger) signed integers. It is the linear conrguential method (x'= ax mod m) with m=2^31-1, the multiplier a=5^7=16807, and coded to avoid integer arithmetic overflow. The choice of the multiplier is the only thing I would question here. The multiplier they give does provide the longest sequence possible for that m, but so does about 534 million other multipliers. So how do we know if 16807 is better than any other multiplier? Knuth provides a test, the Spectral Test, for the "a" used in the linear congruential method to see how good such a generator would be. According to Knuth: "... not only do all good generators pass this test, all generators now known to be bad actually fail it. Thus it is by far the most powerful test known, ...". I ran the spectral test on the multiplier 16807 and also on the multiplier 314159269, which is the only multiplier Knuth ran the spectral test on for m=2^31-1. The results were: a=314159269 a=16807 Minimum for "passing with flying colors" v2^2 1,432,232,969 282,475,250 (bigger is better) v3^2 899,290 408,197 v4^2 36,985 21,682 v5^2 3,427 4,439 v6^2 1,144 895 lg v2 15.2 14.0 15.0 lg v3 9.9 9.3 10.0 lg v4 7.6 7.2 7.5 lg v5 5.9 6.1 6.0 lg v6 5.1 4.9 5.0 u2 2.10 0.41 1.00 u3 1.66 0.51 1.00 u4 3.14 1.08 1.00 u5 1.69 3.22 1.00 u6 3.60 1.73 1.00 So, 16807 is a pretty good multiplier, but it doesn't "pass with flying colors" (based on Knuth's criterion), especially for v2 and u2 which indicate the correlations of any two successive random numbers in the sequence. Since it is common to sample the points in a plane in a simulation or Monte Carlo calculation, this could make the results sensitive to the random number generator. Except for five-dimensional correlations, the multipier 314159269 is clearly better than 16807. Of course, out of the half a billion possible, there may be better multipliers than 314159269, but it would take a while to find one. I use my own random number generators written in assembler that are not only very fast, but also use multipliers with better results than either of those above. However they use m=2^32 and depend on the way multiplication overflows, and so do not lend themselves to portability. I also switch between a linear congruential method and a completely different random number generator (the additive method, which unfortunately cannot be subjected to the spectral test), just to make sure that any calculations I do are not dependent on the random number generator. Since no random number generator can be proved to be good, the idea of a "standard" random number generator for other than teaching purposes seems foolish to me. You would want several so that you can have some confidence that your results are real and not an artifact of the random number generator itself. Mark Adler bitnet: madler@hamlet arpa: madler@hamlet.caltech.edu ------------------------------ Date: Mon, 6-FEB-1989 22:30 +0100 From: "UBMJS2::RMEYER" <U0018%DGOGWDG5.BITNET@CUNYVM.CUNY.EDU> Subject: Hard Disk Info Wanted >The manufacturer is MICROPOLIS, The type is HDA#30059 (at least I found >that number on the case). > >The media defect table - also on the case - reports no less than 27 >errors. Least and highest numbers that occur are > > 0 .. 4 Hd > 3 .. 826 Ctl >and 98 .. 10370 BFIND I found, that a MICROPOLIS 1303 has this Parameters: 5 heads 830 cylinders 17 sectors I think you can use this, on my Commodore PC 60 it is drive type 27. You can use the drive number even if it is not exact the drive you have. I use a CDC WREN 94155-86 (925 cylinders, 9 heads ) as drive type 46 (1024 cylinders, 9 heads). At system start time I get a hard disk failure, but I don't nedd CDCs DMDRVR.BIN ( needs 150k of system memory). Reinhold Meyer Abt. Forstliche Biometrie u. Informatik Buesgenweg 5 D-3400 Goettingen BITNET: U0018@DGOGWDG5 ------------------------------ Date: 5 Feb 89 23:07:00 EST From: <smedleyr@wpafb-info2.arpa> Subject: Hard Disk Interleave I have recently had the opportunity to reset my hard disk interleave factor. I have an old NEC 20MB D3126 3.5 inch (about 80 ms access). The following table applies: Interleave Disk Index (from Norton SI) 2:1 2.3 3:1 1.9 4:1 1.6 6:1 1.3 The last line was the default value of my hard disk for the past three years. I have run the system at 2:1 for the past month with no ill effects. Found similiar results on Seagate ST-251 and ST-8096. There are a couple of packages that will allow you to change the interleave without having to perform a low level format. The only one I can remember off the top of my head is OPTUNE by Gazelle Systems (standard disclaimer). Mark Meaders [Another one that I've used on 'clone' is ILEAVE.arc, available from the archives here at Simtel20 in file: PD1:<MSDOS.DSKUTL>ILEAVE16.ARC (Version I used was ILEAVE15...) gph] ------------------------------ Date: Tue, 7 Feb 1989 01:20 MST From: Keith Petersen <w8sdz@WSMR-SIMTEL20.ARMY.MIL> Subject: Important update to DSZ fixes bug The latest version of DSZ, the x/y/zmodem protocol communications program, is now available via stadnard anonymous FTP from Simtel20. This is a "must get" if you are using the 0117 or 0202 version. It fixes a bug that caused garbled file transfers under some conditions. These files were obtained directly from Chuck Forsberg's BBS just a few hours before this message was posted. Filename Type Bytes CRC Directory PD1:<MSDOS.ZMODEM> DSZ0205X.ARC BINARY 46276 8F6AH <--EXE version DSZ0206.ARC BINARY 81686 FAE0H <--COM version --Keith Petersen Maintainer of the CP/M & MSDOS archives at wsmr-simtel20.army.mil [26.0.0.74] DDN: w8sdz@wsmr-simtel20.army.mil Uucp: {ames,decwrl,harvard,rutgers,ucbvax,uunet}!wsmr-simtel20.army.mil!w8sdz ------------------------------ Date: Mon, 6 Feb 89 10:19:10 EST From: "Nelson R. Pardee" <SASMAINT@SUVM.ACS.SYR.EDU> Subject: I/O on an Epson Equity II Peter DeFrisse writes about the I/O on an Epson Equity II, feeling it is slow due to the Epson DMA. Two responses: -the II is no longer sold. The II+ is a 286 machine. -I have benchmarks for SAS/PC 6.02 on about 25 PC's, including the II. SAS is sensitive to both CPU and disk speed. Time for a vanilla PC with a 65 ms drive was 22:30. The II with the RLL version of the same drive was 13:30. Another clone running at 8 mhz and a similar drive clocked in at about 17. A Zenith 158 running at 8mhz(which I believe doesn't have DMA) clocks in at 21. All those numbers indicate the II has pretty standard performance for a 10 mhz machine. I wonder if something else is wrong. For example, that's the behavior I'd expect if the interleave on the hard drive were set too *low*. We have a couple of II's and have been generally pleased. --Nelson R. Pardee, Academic Computing Services, Syracuse University-- --215 Machinery Hall, Syracuse, NY USA 13244-1260 (315)443-4505 -- ------------------------------ Date: Mon, 6 Feb 89 09:09 MST From: GORDON_A%CUBLDR%VAXF.COLORADO.EDU@CUNYVM.CUNY.EDU Subject: Waveform Editors for PC's regarding waveform editors for PC's: you might try ASYST, which is quite comprehensive. It has reasonably good graphics drivers as well as a complete analysis pkg. It is programmed in a subset of FORTH. You can reach them at 800-348-0033. It is somewhat costly but institutional discounts (25%) are available. Allen Gordon Univ of Colorado ------------------------------ Date: Mon, 06 Feb 89 12:57:33 CST From: David Benson <C09800DB%WUVMD.BITNET@UICVM.uic.edu> Subject: Lotus 123 behaviour problem > My mother is a physician in the Veteran's Administration hospital >system, and has received from them a government-release version of Lotus >1-2-3. This version has a special title-page which explains how it's not >copy-protected, etc. . It sounds like someone has used the Lotus Value Pack to take the copy protection off the Lotus disk. I wonder if they have then made multiple copies? I doubt Lotus would like that much > Now a very odd thing has started happening. After about 5 minutes in >Lotus (sometimes more, sometimes less), whatever one is doing, the screen >suddenly goes blank, a message is posted in the middle saying the drivers' >set file 123.SET can't be found (even though it has to have been loaded >for Lotus to get this far), and when you try to advance beyond the message >Lotus promptly quits. Sounds as though the installation of the Value Pack copy protection removal may have been mis-done. That would be easy since Lotus gives very misleading documentation for it. If this is true, I am not sure there is a recourse. (Has anyone elese had this problem? [Sounds like your mother has a beef with Lotus since the version sold to the Government should not have any sort of 'protection' built in...] - Dave Benson BITNET:C09800DB@WUVMD Associate Director (314) 889-5313 Educational Computing Services Washington University St. Louis, MO Disclaimer: I don't work for Lotus. I am just another of the victims. ------------------------------ Date: Mon, 6-FEB-1989 22:47 +0100 From: "UP200::SCHARFE" <U0158%DGOGWDG5.BITNET@CUNYVM.CUNY.EDU> Subject: millisecond timing on PC Hello, time measurements on a PC is a problem for us, to. We need a millisecond timer for a physical experiment (MICHELSON Interferometer). So I'm very interested, if anyone has a solution for this problem. My suggestion is (on PC/AT): try the "cassette interupt" 15h. There are two nice functions, which may be helpful. INT 15 AH=83h EVENT WAIT AL=0 set interval ES:BX pointer to byte in callers memory that will have the high order bit set as soon as possible after the interval expires CX,DX number of microseconds to elapse befor posting AL=1 Cancel AH=86h WAIT CX,DX number of microseconds to elapse before return to caller (Source: IBM Technical Reference PC/AT, BIOS-Listing page 5-147) But I'm not very happy with this solution, because on most PC's this functions are not implemented correctly (as on our TANDON PCA). For short times (e.g. below a few milliseconds) this method fails. Now I'm looking forward to a better solution on timing problems at the IBM PC. BYE Walter Universitaet Goettingen Walter Scharfe II.Physikalisches Institut Bitnet: U0158@DGOGWDG5 Bunsenstrasse 7-9 D-3400 Goettingen Federal Republic of Germany ------------------------------ Date: Mon, 6 Feb 89 09:02:44 CST From: lantz@skvax1.csc.ti.com Subject: More 4107 questions I have had some response to my question about 4107 emulators and now I have another question. Has anyone used the Grafpoint emulator with a high resolution (1024 X 750 or greater) graphics board in a PC. I have evaluated the package using standard EGA and would like to have some idea how it performs at the higher resolutions. Also, if anyone has a recommendation for a high resolution board, I would like their input. Thanks Bernie Lantz ------------------------------ Date: Sun, 5 Feb 89 18:07 EST From: Richard Ohrbach <ORMOHRBH@ubvmsc.cc.buffalo.edu> Subject: Re: Data acquisition for IBM PS2 model 80 Kumar-- I am not yet using the DT A/D board, but know of others who are using it or similar boards for EMG data collection and analysis. A strong recommendation has been made by them for the software Assyst (or Asyst, I forget the spelling) by a company in Rochester, NY. They make two software programs: Asyst, and Asystant; the latter is easier to use, the former requires programming but provides for more flexibility. The programs are compatible with a variety of A/Ds and preamplifiers. If you are interested in more information on them, I can send you the address and telephone number (I don't have it at home). Richard Ohrbach ------------------------------ Date: Tue, 7 Feb 1989 02:04 MST From: Keith Petersen <w8sdz@WSMR-SIMTEL20.ARMY.MIL> Subject: SIMTEL20 index and CRC list files now available archived The CRC lists and IDX files in Simtel20 top-level directories <CPM>, <MSDOS> and <MISC> are now also available in archive format to save downloading time. Example: Filename Type Bytes CRC Directory PD1:<MSDOS> MSDOSCRC.ARC BINARY 44844 B859H <--MSDOS.CRCLST MSDOSIDX.ARC BINARY 59332 CAD5H <--FILES.IDX The <MISC> and <CPM> filenames are: MISC-CRC.ARC, MISC-IDX.ARC, CPM-CRC.ARC, CPM-IDX.ARC --Keith Petersen Maintainer of the CP/M & MSDOS archives at wsmr-simtel20.army.mil [26.0.0.74] DDN: w8sdz@wsmr-simtel20.army.mil Uucp: {ames,decwrl,harvard,rutgers,ucbvax,uunet}!wsmr-simtel20.army.mil!w8sdz ------------------------------ Date: Mon, 6 Feb 89 08:15 EDT From: <P_GEREMI%UNHH.BITNET@MITVMA.MIT.EDU> Subject: sound for ibm I was just wondering if anyone knows of a digitizing/playback system that can be used on the IBM. I would like to be able to digitize sound and then play it back (to be used in batch files). I was looking for a system which has good quality sound and does not use the speaker system on the IBM. The IBM's sound capabilities are very limited and I was looking for something which could reproduce sound and record sound like the Mac. Any help would be Appreciated. Peter P. Geremia University of New Hampshire IBM Student Representative P_Geremi@UNHH PS, When I try to sell IBM computers at the University Computer store, the mac has the advantage because of the "NEAT" sound. If the IBM had the sound the MAC would not have a chance! ------------------------------ Date: Mon, 6 Feb 89 02:54:02 est From: Lawrence C Foard <lfoard%wpi.wpi.edu@RELAY.CS.NET> Subject: vga pallete Does any one know how to access the VGA pallete directly without using BIOS? I need to be able to use pallete animation and the BIOS routines take to long to change the pallete. By the way does any one know if VEGA VGA has a 640X480 256 color mode, and if so how can it be accessed? ------------------------------ Date: TUE, 7 FEB 89 12:43:49 MEZ From: Raymond Koluvek <R02RDK%DHHDESY3.BITNET@CUNYVM.CUNY.EDU> Subject: X windows (server) for PC I am looking for a distributor of PC Xsight (X windows for a pc) from Locus Computing Corp in Europe, or simular procduct. We have a great number of PC's and would like to try out the concept. Thanks, Ray Raymond D. Koluvek (Deutsches Eletronen - Synchrotron, DESY) Buggy Mail : DESY - Dept. R2, Notkestrasse 85 D-2000 Hamburg 52, West Germany Voice : +49-40-8998-2362 FAX : +49-40-8998-3282 Email : R02RDK@DHHDESY3.BITNET ------------------------------ End of Info-IBMPC Digest ************************ -------