From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: Fri, 21 Dec 01 12:39:03 GMT Organization: UltraNet Communications, Inc. Lines: 46 Message-ID: <9vvhri$k52$10@bob.news.rcn.net> References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> X-Trace: UmFuZG9tSVZB7pXYCTE5osrMxlZp8Z+MhMMBp8qNexBd8Y5GR0RJYXv7g+2aD8Z2 X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 21 Dec 2001 14:44:02 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!feed2.news.rcn.net!feed1.news.rcn.net!rcn!209-122-255-45 Xref: chonsp.franklin.ch alt.sys.pdp10:8399 In article , koehler@encompasserve.org (Bob Koehler) wrote: >In article <85bsgtosjc.fsf_-_@junk.nocrew.org>, Lars Brinkhoff writes: >> koehler@encompasserve.org (Bob Koehler) writes: >>> 1) PDP-10 is the only architecture I've worked with which has no >>> user accessable CPU registers. One tends to use the accumulators >>> on a PDP-10 as one would use registers on other systems, so for >>> this comment I'll use the generic word register for both >> >> How are accumulators different from registers other than the name? > > The accumulators were the first 4 bits of addresss space. Huh? Oh, you don't mean bits...you mean words. > ... You could > put code in them and jump to it, it would execute. You could use an > accumulator address anywhere a memory address was valid. That was a feature. >>> 6) [...] the ability of some instructions to use any accumulator >>> as the stack pointer, while other instruction assumed the >>> standard stack pointer. >> >> I'm not aware of any hardware instruction that assumes a standard >> stack pointer. Maybe there were assembler macros doing that? > > I don't have the reference anymore, but I think it was the JSYS or > some of the JRST. You really are confused. I bet the -10 was the first system you used machine language. Are you talking about PUSHJ/POPJ which do require a stack to be setup once? If you didn't like to set up that stack pointer why didn't you just use the JSR/JSP instructions? /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: Sat, 22 Dec 01 10:29:02 GMT Organization: UltraNet Communications, Inc. Lines: 41 Message-ID: References: <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org>

<3C2385EE.F47B21AF@virgin.net> <3C2395AB.66691CE5@virgin.net> <3C23D858.5FD4AA45@virgin.net> X-Trace: UmFuZG9tSVaq1CWnyBRD2HFrn0diFwb9dVUKxw+aEhEWVDDi35smemfUrhe1T8eJ X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 22 Dec 2001 12:34:12 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!cpk-news-hub1.bbnplanet.com!news.gtei.net!feed2.news.rcn.net!feed1.news.rcn.net!rcn!207-172-216-167 Xref: chonsp.franklin.ch alt.sys.pdp10:8569 In article <3C23D858.5FD4AA45@virgin.net>, Alan Greig wrote: > > >Christopher Stacy wrote: > >> >>>>> On Fri, 21 Dec 2001 20:03:55 +0000, Alan Greig ("Alan") writes: >> Alan> Alan Greig wrote: >> >> Ah wait I think Bob might also be thinking of the NOOP JUMPA that wasn't a >> >> That stuff was not a feature of the PDP-10, but rather a feature of TOPS-20. > >I know. Next you'll be telling me that jsys% was a feature of TOPS-20 rather than >the PDP-10 :-) > >Little known fact (well maybe not to some people here!). TOPS-20 implemented part >of one TOPS-10 UUO entirely natively. A GETTAB UUO for the OPSYS type was >understood directly by TOPS-20 monitor. Control was not passed to the PA-1050 >TOPS-10 emulator. Even the TOPS-10 Monitor calls manual gave the value returned >for TOPS-20. > >In theory an executable could figure out whether it was running under TOPS-10 or >TOPS-20 and use native calls accordingly. Not that I recall any that ever did. Yup. There was a standard. Each OS was assigned a value. I can't remember the symbols in UUOSYM but they're there...IIRC, even ITS. If an entity outside of DEC wished to write a new OS and have it assigned a value for UUO interrogation purposes, there was a procedure in place to assign a value for that non-DEC entity. /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: Sat, 22 Dec 01 10:31:04 GMT Organization: UltraNet Communications, Inc. Lines: 21 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> <9vvhri$k52$10@bob.news.rcn.net> X-Trace: UmFuZG9tSVY3qKffy2PYLWpzUF5ZmpqrUK0tb0NT4N6fr3pKekWWx3NAQ1lQUKlr X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 22 Dec 2001 12:36:13 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!cpk-news-hub1.bbnplanet.com!news.gtei.net!feed2.news.rcn.net!feed1.news.rcn.net!rcn!207-172-216-167 Xref: chonsp.franklin.ch alt.sys.pdp10:8574 In article , Jan Vorbrueggen wrote: >jmfbahciv@aol.com writes: > >> > ... You could >> > put code in them and jump to it, it would execute. You could use an >> > accumulator address anywhere a memory address was valid. >> That was a feature. > >....that would severely limit a modern implementation of >the architecture. And >void any advantage an I-cache might bring. If you had a machine so broken that you needed to run in the accumulators, I'm pretty sure that the caches would be turned off. Zeroing memory on a cold boot was a very nice feature. /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: Sat, 22 Dec 01 10:34:04 GMT Organization: UltraNet Communications, Inc. Lines: 59 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> <3C237513.B0746365@virgin.net> X-Trace: UmFuZG9tSVaBuZvsDW1oNci0EQOSqJywcSsYcs8jH5w0+on96XC1UsliD5vQxCGr X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 22 Dec 2001 12:39:14 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!feed2.news.rcn.net!feed1.news.rcn.net!rcn!207-172-216-167 Xref: chonsp.franklin.ch alt.sys.pdp10:8573 In article <3C237513.B0746365@virgin.net>, Alan Greig wrote: > > >Bob Koehler wrote: > >> I >> >> The accumulators were the first 4 bits of addresss space. You could >> put code in them and jump to it, it would execute. You could use an >> accumulator address anywhere a memory address was valid. >> > >And very handy that could be as well. A program could copy a small section of itself into the >accumulators and then jump into them. That code could load another program (say) into the same >address space. Fast tight loops could also execute in the accumulators although this was more of >a benefit on the earlier pre KL-10 PDP-10s > >> I don't have the reference anymore, but I think it was the JSYS or >> some of the JRST. >> > >Ah sounds like you are thinking of pre-defined MACROs such as ERMSG and ERHLT which I think >expanded to something like: > >ERMSG < Unexpected error -> > > PUSHJ 17, [hrroi 1,[asciz/Unexpected Error/ > psout% > setom 1 ; -1 last error code > esout% ; output error strinng > popj 17] ; continue execution > >Probably have the argument to esout% slightly wrong as I'm not sure it went to .priou (primary >output - sys$output) by default. psout% - primary output string out did. > >Yes these predefined expansions assumed a stack pointer in 17 (octal) but nothing forced you to >use them. > > I always wondered if those MACROes would be misinterpreted for machine instructions. (It was one of the reasons I didn't like them.) Now I know. I was right yet again. One of the features of generating a MACRO listing was that those macroes were expanded so that you could tell if the fucking thing was a macro (look at the machine instructions that occurred on the left hand side of the page). Didn't any CS classes study a CREF listing? /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: Sat, 22 Dec 01 10:35:41 GMT Organization: UltraNet Communications, Inc. Lines: 32 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <85bsgtosjc.fsf_-_@junk.nocrew.org> <3C237513.B0746365@virgin.net> X-Trace: UmFuZG9tSVbuceaBUp4Ao/3fSZ94Eym9SltKZVkrjunSFsrKXJYnJJFuZAs84Gto X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 22 Dec 2001 12:40:51 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!feed2.news.rcn.net!feed1.news.rcn.net!rcn!207-172-216-167 Xref: chonsp.franklin.ch alt.sys.pdp10:8568 In article , peter@taronga.com (Peter da Silva) wrote: >In article <3C237513.B0746365@virgin.net>, >Alan Greig wrote: >>And very handy that could be as well. A program could copy a small >>section of itself into the >>accumulators and then jump into them. That code could load another >>program (say) into the same >>address space. Fast tight loops could also execute in the accumulators >>although this was more of >>a benefit on the earlier pre KL-10 PDP-10s > ><------------- please adjust your window to this width ------ :) --------> > >Once you have cache, though, this doesn't save you anything and has the >disadvantage of making it very difficult to pipeline accesses to the >accumulators, and someone else has already mentioned what this does to >separate I&D caching. > >Not that the vax is particularly easy to pipeline. You need just-in-time >translation techniques to make the VAX run really fast. And they became >really practical far too late for the VAX, though I understand DEC came >up with a nice software JIT engine for the VAX-Alpha transition. > no, no, no. You don't get the point. If you have to run in the accumulators, you aren't running the system in all its glory. It's either very broken or every little bit of address space has to be free for the program. /BAH Subtract a hundred and four for e-mail. ###### From: Lars Brinkhoff Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 21 Dec 2001 10:01:59 +0100 Organization: nocrew Lines: 52 Message-ID: <85bsgtosjc.fsf_-_@junk.nocrew.org> References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> NNTP-Posting-Host: junk.nocrew.org Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Trace: yggdrasil.utfors.se 1008925298 5644 212.73.17.42 (21 Dec 2001 09:01:38 GMT) X-Complaints-To: abuse@utfors.se NNTP-Posting-Date: 21 Dec 2001 09:01:38 GMT User-Agent: Gnus/5.0808 (Gnus v5.8.8) Emacs/20.7 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.tele.dk!small.news.tele.dk!212.37.1.10!newsfeed01.nntp.se.dataphone.net!nntp.se.dataphone.net!newsfeed.sunet.se!news01.sunet.se!news.utfors.se!unknown!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8458 koehler@encompasserve.org (Bob Koehler) writes: > 1) PDP-10 is the only architecture I've worked with which has no > user accessable CPU registers. One tends to use the accumulators > on a PDP-10 as one would use registers on other systems, so for > this comment I'll use the generic word register for both How are accumulators different from registers other than the name? > the PDP-10 is the only system I've used on which one register > (accumulator 0) is sometimes a real register and sometimes > ignored. I think the ADJBP instruction would be an example of this. > 2) The use of a conditional skip instead of a conditional branch. There are six conditional JUMP instructions. > 4) The limitted address range, even extended addressing was > limitted to 23 bits. True for the KL10 implementation. Jupiter would have 30 bits, XKL's TOAD-1 has 30 bits, and I think the SC-40 has (had?) 30 bits as well. > 5) The requirement (except on the 2020) to use a PDP-11 in order > to communicate with the outside world. I don't think this is a requirement inherent in the PDP-10 architecture? > 6) [...] the ability of some instructions to use any accumulator > as the stack pointer, while other instruction assumed the > standard stack pointer. I'm not aware of any hardware instruction that assumes a standard stack pointer. Maybe there were assembler macros doing that? > 7) Not enough accumulators. There are 16 (or 15 if you don't consider accumulator 0 to be general purpose). I'll note that at least a modern RISC architecture, namely ARM, has 16 registers, and seems to be doing well. > 8) No support for paging (virtual memory concepts were limitted > to swapping). All PDP-10 implementations since the KI10 had support for paging. Even the original (KA10) PDP-10 was commonly used with an external pager (MMU). -- Lars Brinkhoff http://lars.nocrew.org/ Linux, GCC, PDP-10 Brinkhoff Consulting http://www.brinkhoff.se/ programming ###### From: gah@ugcs.caltech.edu (glen herrmannsfeldt) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 21 Dec 2001 09:51:13 GMT Organization: California Institute of Technology, Pasadena Lines: 22 Message-ID: <9vv0mh$lu3@gap.cco.caltech.edu> References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> NNTP-Posting-Host: seniti.ugcs.caltech.edu X-Newsreader: NN version 6.5.0 #1 (NOV) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!logbridge.uoregon.edu!nntp-server.caltech.edu!gah Xref: chonsp.franklin.ch alt.sys.pdp10:8425 koehler@encompasserve.org (Bob Koehler) writes: (snip) > the PDP-10 is the only system I've used on which one register > (accumulator 0) is sometimes a real register and sometimes > ignored. This is very common. S/360, S/370, etc., and mose RISC machines. Maybe it is uncommon not to do it on a multi-register machine. (snip) > 4) The limitted address range, even extended addressing was > limitted to 23 bits. I think address space is the real limitation on the architecture. Maybe a compatible extension could have been done, though. (snip) > 7) Not enough accumulators. S/390 is still going with its original 16. It is a little hard on a compiler, but it isn't so bad. Good hardware design can work around even the more limited x86 register set. -- glen ###### From: koehler@encompasserve.org (Bob Koehler) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 21 Dec 2001 08:30:00 -0600 Organization: Encompasserve Lines: 61 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> NNTP-Posting-Host: eisner.encompasserve.org X-Trace: grandcanyon.binc.net 1008945003 5464 192.135.80.34 (21 Dec 2001 14:30:03 GMT) X-Complaints-To: abuse@binc.net NNTP-Posting-Date: Fri, 21 Dec 2001 14:30:03 +0000 (UTC) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.cwix.com!news.binc.net!koehler Xref: chonsp.franklin.ch alt.sys.pdp10:8494 In article <85bsgtosjc.fsf_-_@junk.nocrew.org>, Lars Brinkhoff writes: > koehler@encompasserve.org (Bob Koehler) writes: >> 1) PDP-10 is the only architecture I've worked with which has no >> user accessable CPU registers. One tends to use the accumulators >> on a PDP-10 as one would use registers on other systems, so for >> this comment I'll use the generic word register for both > > How are accumulators different from registers other than the name? The accumulators were the first 4 bits of addresss space. You could put code in them and jump to it, it would execute. You could use an accumulator address anywhere a memory address was valid. > >> the PDP-10 is the only system I've used on which one register >> (accumulator 0) is sometimes a real register and sometimes >> ignored. > > I think the ADJBP instruction would be an example of this. > >> 2) The use of a conditional skip instead of a conditional branch. > > There are six conditional JUMP instructions. > >> 4) The limitted address range, even extended addressing was >> limitted to 23 bits. > > True for the KL10 implementation. Jupiter would have 30 bits, XKL's > TOAD-1 has 30 bits, and I think the SC-40 has (had?) 30 bits as well. > >> 5) The requirement (except on the 2020) to use a PDP-11 in order >> to communicate with the outside world. > > I don't think this is a requirement inherent in the PDP-10 architecture? > >> 6) [...] the ability of some instructions to use any accumulator >> as the stack pointer, while other instruction assumed the >> standard stack pointer. > > I'm not aware of any hardware instruction that assumes a standard > stack pointer. Maybe there were assembler macros doing that? I don't have the reference anymore, but I think it was the JSYS or some of the JRST. >> 7) Not enough accumulators. > > There are 16 (or 15 if you don't consider accumulator 0 to be general > purpose). I'll note that at least a modern RISC architecture, namely > ARM, has 16 registers, and seems to be doing well. > >> 8) No support for paging (virtual memory concepts were limitted >> to swapping). > > All PDP-10 implementations since the KI10 had support for paging. > Even the original (KA10) PDP-10 was commonly used with an external > pager (MMU). > > -- > Lars Brinkhoff http://lars.nocrew.org/ Linux, GCC, PDP-10 > Brinkhoff Consulting http://www.brinkhoff.se/ programming ###### From: Jan Vorbrueggen Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 21 Dec 2001 18:04:04 +0100 Organization: Institut fuer Neuroinformatik, Ruhr-Universitaet Bochum, Germany Lines: 11 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> <9vvhri$k52$10@bob.news.rcn.net> NNTP-Posting-Host: luda.neuroinformatik.ruhr-uni-bochum.de X-Trace: sunu789.rz.ruhr-uni-bochum.de 1008954245 29714 134.147.176.178 (21 Dec 2001 17:04:05 GMT) NNTP-Posting-Date: 21 Dec 2001 17:04:05 GMT X-Newsreader: Gnus v5.7/Emacs 20.5 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!news.imp.ch!uni-erlangen.de!news-nue1.dfn.de!news-han1.dfn.de!news-koe1.dfn.de!news.ruhr-uni-bochum.de!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8506 jmfbahciv@aol.com writes: > > ... You could > > put code in them and jump to it, it would execute. You could use an > > accumulator address anywhere a memory address was valid. > That was a feature. ...that would severely limit a modern implementation of the architecture. And void any advantage an I-cache might bring. Jan ###### Message-ID: <3C237513.B0746365@virgin.net> From: Alan Greig X-Mailer: Mozilla 4.6 [en] (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 42 Date: Fri, 21 Dec 2001 17:44:51 +0000 NNTP-Posting-Host: 62.252.135.22 X-Complaints-To: abuse@virgin.net X-Trace: news6-win.server.ntlworld.com 1008956719 62.252.135.22 (Fri, 21 Dec 2001 17:45:19 GMT) NNTP-Posting-Date: Fri, 21 Dec 2001 17:45:19 GMT Organization: Virgin Net Usenet Service Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed00.sul.t-online.de!t-online.de!diablo.theplanet.net!newspeer.clara.net!news.clara.net!news5-gui.server.ntli.net!ntli.net!news6-win.server.ntlworld.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8347 Bob Koehler wrote: > I > > The accumulators were the first 4 bits of addresss space. You could > put code in them and jump to it, it would execute. You could use an > accumulator address anywhere a memory address was valid. > And very handy that could be as well. A program could copy a small section of itself into the accumulators and then jump into them. That code could load another program (say) into the same address space. Fast tight loops could also execute in the accumulators although this was more of a benefit on the earlier pre KL-10 PDP-10s > I don't have the reference anymore, but I think it was the JSYS or > some of the JRST. > Ah sounds like you are thinking of pre-defined MACROs such as ERMSG and ERHLT which I think expanded to something like: ERMSG < Unexpected error -> PUSHJ 17, [hrroi 1,[asciz/Unexpected Error/ psout% setom 1 ; -1 last error code esout% ; output error strinng popj 17] ; continue execution Probably have the argument to esout% slightly wrong as I'm not sure it went to .priou (primary output - sys$output) by default. psout% - primary output string out did. Yes these predefined expansions assumed a stack pointer in 17 (octal) but nothing forced you to use them. -- Alan Greig ###### Message-ID: <3C2385EE.F47B21AF@virgin.net> From: Alan Greig X-Mailer: Mozilla 4.6 [en] (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org>

Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 42 Date: Fri, 21 Dec 2001 18:56:46 +0000 NNTP-Posting-Host: 62.252.135.22 X-Complaints-To: abuse@virgin.net X-Trace: news2-win.server.ntlworld.com 1008961025 62.252.135.22 (Fri, 21 Dec 2001 18:57:05 GMT) NNTP-Posting-Date: Fri, 21 Dec 2001 18:57:05 GMT Organization: Virgin Net Usenet Service Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!skynet.be!skynet.be!tank.news.pipex.net!pipex!newspeer.clara.net!news.clara.net!news5-gui.server.ntli.net!ntli.net!news2-win.server.ntlworld.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8361 Mark Crispin wrote: > On 21 Dec 2001, Bob Koehler wrote: > > >> 6) [...] the ability of some instructions to use any accumulator > > >> as the stack pointer, while other instruction assumed the > > >> standard stack pointer. > > > I'm not aware of any hardware instruction that assumes a standard > > > stack pointer. Maybe there were assembler macros doing that? > > I don't have the reference anymore, but I think it was the JSYS or > > some of the JRST. > > Try "Bob Koehler's fevered imagination". > > There were no machine instructions that made any assumptions about the > stack pointer. > > Nor were there any system calls (JSYS or UUO) on any operating system > which made any such assumptions. > > The only thing that assumed AC 17 were some macros in the MACSYM package, > usage of which was totally optional (and was specific to TOPS-20). > Ah wait I think Bob might also be thinking of the NOOP JUMPA that wasn't a NOOP if the *immediately* preceding JSYS call failed but still had a +1PC return. JUMPA with a particular AC wasn't it? Was their a call as well? ERJMP and ERCALL macros? Once upon a time I wouldn't have even had to think about this. Really need to upgrade my home PC to something that can reasonably run the emulators as I can't justify much time at work. > > -- Mark -- > > http://staff.washington.edu/mrc > Science does not emerge from voting, party politics, or public debate. -- Alan Greig ###### From: peter@taronga.com (Peter da Silva) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 21 Dec 2001 19:36:10 GMT Organization: TSS Inc. Lines: 27 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <85bsgtosjc.fsf_-_@junk.nocrew.org> <3C237513.B0746365@virgin.net> NNTP-Posting-Host: citadel.in.taronga.com X-Trace: citadel.in.taronga.com 1008963370 40538 10.0.0.43 (21 Dec 2001 19:36:10 GMT) X-Complaints-To: usenet@taronga.com NNTP-Posting-Date: 21 Dec 2001 19:36:10 GMT X-Newsreader: trn 4.0-test72 (19 April 1999) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.stanford.edu!news-proxy.baileynm.com!citadel.in.taronga.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8449 In article <3C237513.B0746365@virgin.net>, Alan Greig wrote: >And very handy that could be as well. A program could copy a small >section of itself into the >accumulators and then jump into them. That code could load another >program (say) into the same >address space. Fast tight loops could also execute in the accumulators >although this was more of >a benefit on the earlier pre KL-10 PDP-10s <------------- please adjust your window to this width ------ :) --------> Once you have cache, though, this doesn't save you anything and has the disadvantage of making it very difficult to pipeline accesses to the accumulators, and someone else has already mentioned what this does to separate I&D caching. Not that the vax is particularly easy to pipeline. You need just-in-time translation techniques to make the VAX run really fast. And they became really practical far too late for the VAX, though I understand DEC came up with a nice software JIT engine for the VAX-Alpha transition. -- Rev. Peter da Silva, ULC. "Cave cuniculos lagana ferentes" "Be conservative in what you generate, and liberal in what you accept" -- Matthew 10:16 (l.trans) ###### Message-ID: <3C2395AB.66691CE5@virgin.net> From: Alan Greig X-Mailer: Mozilla 4.6 [en] (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org>

<3C2385EE.F47B21AF@virgin.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 17 Date: Fri, 21 Dec 2001 20:03:55 +0000 NNTP-Posting-Host: 62.252.132.28 X-Complaints-To: abuse@virgin.net X-Trace: news2-win.server.ntlworld.com 1008965055 62.252.132.28 (Fri, 21 Dec 2001 20:04:15 GMT) NNTP-Posting-Date: Fri, 21 Dec 2001 20:04:15 GMT Organization: Virgin Net Usenet Service Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newscore.univie.ac.at!194.74.65.73.MISMATCH!btnet-peer0!btnet!news5-gui.server.ntli.net!ntli.net!news2-win.server.ntlworld.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8354 Alan Greig wrote: > > > Ah wait I think Bob might also be thinking of the NOOP JUMPA that wasn't a > On more reflection a *fatal* JSYS error could be trapped if the next instruction was a certain noop JUMPA. JSHLT% macro? Yes/no? Think my brain's currently running some complex ECC algorithms on old memory... -- Alan Greig ###### From: koehler@encompasserve.org (Bob Koehler) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 21 Dec 2001 14:05:53 -0600 Organization: Encompasserve Lines: 8 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> <9vvhri$k52$10@bob.news.rcn.net> NNTP-Posting-Host: eisner.encompasserve.org X-Trace: grandcanyon.binc.net 1008965154 11205 192.135.80.34 (21 Dec 2001 20:05:54 GMT) X-Complaints-To: abuse@binc.net NNTP-Posting-Date: Fri, 21 Dec 2001 20:05:54 +0000 (UTC) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.cwix.com!news.binc.net!koehler Xref: chonsp.franklin.ch alt.sys.pdp10:8501 In article <9vvhri$k52$10@bob.news.rcn.net>, jmfbahciv@aol.com writes: > > Huh? Oh, you don't mean bits...you mean words. > The first 16 (decimal) words. 0 through 17 in octal. 0000 though 1111 in binary. 4 bits worth of address space. ###### Message-ID: <3C239703.B0DF989@virgin.net> From: Alan Greig X-Mailer: Mozilla 4.6 [en] (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <85bsgtosjc.fsf_-_@junk.nocrew.org> <3C237513.B0746365@virgin.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 40 Date: Fri, 21 Dec 2001 20:09:39 +0000 NNTP-Posting-Host: 62.252.132.28 X-Complaints-To: abuse@virgin.net X-Trace: news2-win.server.ntlworld.com 1008965398 62.252.132.28 (Fri, 21 Dec 2001 20:09:58 GMT) NNTP-Posting-Date: Fri, 21 Dec 2001 20:09:58 GMT Organization: Virgin Net Usenet Service Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!skynet.be!skynet.be!newspeer.clara.net!news.clara.net!news5-gui.server.ntli.net!ntli.net!news2-win.server.ntlworld.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8345 Peter da Silva wrote: > <------------- please adjust your window to this width ------ :) --------> > The version of Netscape I post from home with seems to have a mind of its own. Some posts are ok but others aren't. Short of re-installing (the typical Windows world fix) I'm a bit stumped. > > Once you have cache, though, this doesn't save you anything and has the > disadvantage of making it very difficult to pipeline accesses to the > accumulators, and someone else has already mentioned what this does to > separate I&D caching. > For the special case of executing code in the ACs I guess a modern architecture could work around the optimization problem without too much difficulty. After all we are only talking of this applying to the first 16 words in address space. > > Not that the vax is particularly easy to pipeline. You need just-in-time > translation techniques to make the VAX run really fast. And they became > really practical far too late for the VAX, though I understand DEC came > up with a nice software JIT engine for the VAX-Alpha transition. > > -- > Rev. Peter da Silva, ULC. "Cave cuniculos lagana ferentes" > > "Be conservative in what you generate, and liberal in what you accept" > -- Matthew 10:16 (l.trans) -- Alan Greig ###### Message-ID: <3C23D858.5FD4AA45@virgin.net> From: Alan Greig X-Mailer: Mozilla 4.6 [en] (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org>

<3C2385EE.F47B21AF@virgin.net> <3C2395AB.66691CE5@virgin.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 25 Date: Sat, 22 Dec 2001 00:48:25 +0000 NNTP-Posting-Host: 62.252.132.33 X-Complaints-To: abuse@virgin.net X-Trace: news6-win.server.ntlworld.com 1008982124 62.252.132.33 (Sat, 22 Dec 2001 00:48:44 GMT) NNTP-Posting-Date: Sat, 22 Dec 2001 00:48:44 GMT Organization: Virgin Net Usenet Service Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed00.sul.t-online.de!t-online.de!grolier!btnet-peer0!btnet!news5-gui.server.ntli.net!ntli.net!news6-win.server.ntlworld.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8551 Christopher Stacy wrote: > >>>>> On Fri, 21 Dec 2001 20:03:55 +0000, Alan Greig ("Alan") writes: > Alan> Alan Greig wrote: > >> Ah wait I think Bob might also be thinking of the NOOP JUMPA that wasn't a > > That stuff was not a feature of the PDP-10, but rather a feature of TOPS-20. I know. Next you'll be telling me that jsys% was a feature of TOPS-20 rather than the PDP-10 :-) Little known fact (well maybe not to some people here!). TOPS-20 implemented part of one TOPS-10 UUO entirely natively. A GETTAB UUO for the OPSYS type was understood directly by TOPS-20 monitor. Control was not passed to the PA-1050 TOPS-10 emulator. Even the TOPS-10 Monitor calls manual gave the value returned for TOPS-20. In theory an executable could figure out whether it was running under TOPS-10 or TOPS-20 and use native calls accordingly. Not that I recall any that ever did. -- Alan Greig ###### From: gah@ugcs.caltech.edu (glen herrmannsfeldt) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 22 Dec 2001 18:28:11 GMT Organization: California Institute of Technology, Pasadena Lines: 12 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <85bsgtosjc.fsf_-_@junk.nocrew.org> <3C237513.B0746365@virgin.net> NNTP-Posting-Host: yak.ugcs.caltech.edu X-Newsreader: NN version 6.5.0 #1 (NOV) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!logbridge.uoregon.edu!nntp-server.caltech.edu!gah Xref: chonsp.franklin.ch alt.sys.pdp10:8575 jmfbahciv@aol.com writes: (snip) >no, no, no. You don't get the point. If you have to run in >the accumulators, you aren't running the system in all its >glory. It's either very broken or every little bit of address >space has to be free for the program. I understood that the search function in TECO did it. I don't know that I ever heard of anything else. I believe it makes a significant difference on the KA-10, maybe not on others. -- glen ###### Path: chonsp.franklin.ch!not-for-mail From: Neil Franklin Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 22 Dec 2001 23:38:28 +0100 Organization: My own Private Self Lines: 41 Message-ID: <6u7kreaniz.fsf@chonsp.franklin.ch> References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> <9vvhri$k52$10@bob.news.rcn.net> NNTP-Posting-Host: chonsp.franklin.ch X-Trace: chonsp.franklin.ch 1009060710 1504 10.0.3.2 (22 Dec 2001 22:38:30 GMT) X-Complaints-To: news@chonsp.franklin.ch NNTP-Posting-Date: 22 Dec 2001 22:38:30 GMT X-Newsreader: Gnus v5.7/Emacs 20.4 Xref: chonsp.franklin.ch alt.sys.pdp10:8618 Jan Vorbrueggen writes: > jmfbahciv@aol.com writes: > > > > ... You could > > > put code in them and jump to it, it would execute. You could use an > > > accumulator address anywhere a memory address was valid. > > That was a feature. > > ...that would severely limit a modern implementation of the architecture. And > void any advantage an I-cache might bring. Why would that limit it? And why void the I-cache? Just regard the PDP-10 as having an main memory with an fixed-address (from 0) 16 word combined-I/D cache. Then treat the instructions as 2-register + 1-memory address, with the AC and X fields directly reading the accumulators/registers (2 read ports) and Y as 18bit in-segment offset into 18/26/30bit memory space (mapping to 18/22 bit physical memory). For the special case of Y=0..15 have an Mux switch memory input to from an additional register read port. Place your I-cache either between memory and reg/mem Mux (do not cache from registers, as they deliver fast anyway) or after the Mux, treating registers simply as fast cache load (gives more uniform from-cache read). No reason why this should slow down the machine at all. I have programmed x86 and 86k, read a lot on ARM, MIPS, Sparc and PPC implementation, and am presently designing an PDP-10 microprocessor just for the fun of it. I see no reason why I can not achieve 1990 ARM/MIPS/Sparc style performance, from an 1990-chip-size-equvalent FPGA, using an 3-read-port register set. -- Neil Franklin, neil@franklin.ch.remove http://neil.franklin.ch/ Hacker, Unix Guru, El Eng HTL/BSc, Sysadmin, Archer, Roleplayer - Intellectual Property is Intellectual Robbery ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: Sun, 23 Dec 01 10:05:22 GMT Organization: UltraNet Communications, Inc. Lines: 22 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <85bsgtosjc.fsf_-_@junk.nocrew.org> <3C237513.B0746365@virgin.net> X-Trace: UmFuZG9tSVazmGX0y1tHwtfLZCVKGkQFhmrGtsCZ8+egSLfVmuiOkIKrxOUMDXLs X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 23 Dec 2001 12:10:42 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!feed2.news.rcn.net!feed1.news.rcn.net!rcn!207-172-216-133 Xref: chonsp.franklin.ch alt.sys.pdp10:8629 In article , gah@ugcs.caltech.edu (glen herrmannsfeldt) wrote: >jmfbahciv@aol.com writes: >(snip) >>no, no, no. You don't get the point. If you have to run in >>the accumulators, you aren't running the system in all its >>glory. It's either very broken or every little bit of address >>space has to be free for the program. > >I understood that the search function in TECO did it. I don't >know that I ever heard of anything else. I believe it makes >a significant difference on the KA-10, maybe not on others. Oh, I'm sure lots of programs took advantage of that. I'm talking to the people who have given me the distinct idea that they believe hardware can never fuck up. The attitude is understandable in this PCentric world since it's a plug and throw away set of hardware. /BAH Subtract a hundred and four for e-mail. ###### From: Ben Franchuk Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: Sat, 22 Dec 2001 01:22:10 -0700 Organization: Posted via Supernews, http://www.supernews.com Message-ID: <3C2442B2.20D5CE1E@jetnet.ab.ca> X-Mailer: Mozilla 4.79 [en] (Win95; U) X-Accept-Language: en MIME-Version: 1.0 References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> <9vvhri$k52$10@bob.news.rcn.net> <6u7kreaniz.fsf@chonsp.franklin.ch> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Complaints-To: newsabuse@supernews.com Lines: 44 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.stanford.edu!sn-xit-01!sn-post-01!supernews.com!corp.supernews.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8625 Neil Franklin wrote: > > Jan Vorbrueggen writes: > > > jmfbahciv@aol.com writes: > > > > > > ... You could > > > > put code in them and jump to it, it would execute. You could use an > > > > accumulator address anywhere a memory address was valid. > > > That was a feature. > > > > ...that would severely limit a modern implementation of the architecture. And > > void any advantage an I-cache might bring. > > Why would that limit it? And why void the I-cache? > > Just regard the PDP-10 as having an main memory with an fixed-address > (from 0) 16 word combined-I/D cache. > > Then treat the instructions as 2-register + 1-memory address, with the > AC and X fields directly reading the accumulators/registers (2 read > ports) and Y as 18bit in-segment offset into 18/26/30bit memory space > (mapping to 18/22 bit physical memory). > > For the special case of Y=0..15 have an Mux switch memory input to > from an additional register read port. Place your I-cache either > between memory and reg/mem Mux (do not cache from registers, as they > deliver fast anyway) or after the Mux, treating registers simply as > fast cache load (gives more uniform from-cache read). > > No reason why this should slow down the machine at all. > > I have programmed x86 and 86k, read a lot on ARM, MIPS, Sparc and PPC > implementation, and am presently designing an PDP-10 microprocessor > just for the fun of it. I see no reason why I can not achieve 1990 > ARM/MIPS/Sparc style performance, from an 1990-chip-size-equvalent > FPGA, using an 3-read-port register set. > Sounds fast! Just how big is that FPGA to be 1990 equvalent? -- Ben Franchuk --- Pre-historic Cpu's -- www.jetnet.ab.ca/users/bfranchuk/index.html ###### From: peter@taronga.com (Peter da Silva) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 22 Dec 2001 22:52:59 GMT Organization: TSS Inc. Lines: 30 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <3C237513.B0746365@virgin.net> <3C239703.B0DF989@virgin.net> NNTP-Posting-Host: citadel.in.taronga.com X-Trace: citadel.in.taronga.com 1009061579 94523 10.0.0.43 (22 Dec 2001 22:52:59 GMT) X-Complaints-To: usenet@taronga.com NNTP-Posting-Date: 22 Dec 2001 22:52:59 GMT X-Newsreader: trn 4.0-test72 (19 April 1999) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.stanford.edu!news.kjsl.com!news.usenet2.org!citadel.in.taronga.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8656 In article <3C239703.B0DF989@virgin.net>, Alan Greig wrote: >Peter da Silva wrote: >> <------------- please adjust your window to this width ------ :) --------> >The version of Netscape I post from home with seems to have a mind of its own. >Some posts are ok but others aren't. Short of re-installing (the typical >Windows world fix) I'm a bit stumped. I may actually be able to help here. Check your message composition settings (I don't recall where they are, but the Netscape repferences are fairly easy to deal with in both GUI and text forms) and tell it to always use plain text rather than HTML. >For the special case of executing code in the ACs I guess a modern >architecture could work around the optimization problem without too much >difficulty. After all we are only talking of this applying to the first 16 >words in address space. Yeh, but any machine modern enough to do that would be running loops that short in the decoded instruction buffer anyway, and it'd need to add a bunch of logic in the instruction-execuition critical path to make sure it only clobbered the decoded instruction buffer on accumulator writes to code that happened to already be in that buffer. -- Rev. Peter da Silva, ULC. "Cave cuniculos lagana ferentes" "Be conservative in what you generate, and liberal in what you accept" -- Matthew 10:16 (l.trans) ###### Message-ID: <3C251360.3EC78DF4@virgin.net> From: Alan Greig X-Mailer: Mozilla 4.6 [en] (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <3C237513.B0746365@virgin.net> <3C239703.B0DF989@virgin.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 24 Date: Sat, 22 Dec 2001 23:12:32 +0000 NNTP-Posting-Host: 62.252.133.91 X-Complaints-To: abuse@virgin.net X-Trace: news6-win.server.ntlworld.com 1009062772 62.252.133.91 (Sat, 22 Dec 2001 23:12:52 GMT) NNTP-Posting-Date: Sat, 22 Dec 2001 23:12:52 GMT Organization: Virgin Net Usenet Service Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!1917477!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!newsgate.cistron.nl!news2.euro.net!newspeer.clara.net!news.clara.net!news5-gui.server.ntli.net!ntli.net!news6-win.server.ntlworld.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8621 Peter da Silva wrote: > In article <3C239703.B0DF989@virgin.net>, > Alan Greig wrote: > >Peter da Silva wrote: > >> <------------- please adjust your window to this width ------ :) --------> > > >The version of Netscape I post from home with seems to have a mind of its own. > >Some posts are ok but others aren't. Short of re-installing (the typical > >Windows world fix) I'm a bit stumped. > > I may actually be able to help here. Check your message composition settings > (I don't recall where they are, but the Netscape repferences are fairly easy > to deal with in both GUI and text forms) and tell it to always use plain text > rather than HTML. > Already set... -- Alan Greig ###### From: peter@taronga.com (Peter da Silva) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 23 Dec 2001 02:39:34 GMT Organization: TSS Inc. Lines: 16 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <3C239703.B0DF989@virgin.net> <3C251360.3EC78DF4@virgin.net> NNTP-Posting-Host: citadel.in.taronga.com X-Trace: citadel.in.taronga.com 1009075174 1273 10.0.0.43 (23 Dec 2001 02:39:34 GMT) X-Complaints-To: usenet@taronga.com NNTP-Posting-Date: 23 Dec 2001 02:39:34 GMT X-Newsreader: trn 4.0-test72 (19 April 1999) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.stanford.edu!news-proxy.baileynm.com!citadel.in.taronga.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8648 In article <3C251360.3EC78DF4@virgin.net>, Alan Greig wrote: >> I may actually be able to help here. Check your message composition settings >> (I don't recall where they are, but the Netscape repferences are fairly easy >> to deal with in both GUI and text forms) and tell it to always use plain text >> rather than HTML. >Already set... Plan B: install a real operating system and use a real newsreader. -- Rev. Peter da Silva, ULC. "Cave cuniculos lagana ferentes" "Be conservative in what you generate, and liberal in what you accept" -- Matthew 10:16 (l.trans) ###### Message-ID: <3C25947A.7FF3A951@virgin.net> From: Alan Greig X-Mailer: Mozilla 4.6 [en] (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <3C239703.B0DF989@virgin.net> <3C251360.3EC78DF4@virgin.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 40 Date: Sun, 23 Dec 2001 08:23:22 +0000 NNTP-Posting-Host: 62.252.132.10 X-Complaints-To: abuse@virgin.net X-Trace: news2-win.server.ntlworld.com 1009095824 62.252.132.10 (Sun, 23 Dec 2001 08:23:44 GMT) NNTP-Posting-Date: Sun, 23 Dec 2001 08:23:44 GMT Organization: Virgin Net Usenet Service Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!skynet.be!skynet.be!dispose.news.demon.net!demon!btnet-peer0!btnet!news5-gui.server.ntli.net!ntli.net!news2-win.server.ntlworld.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8622 Peter da Silva wrote: > In article <3C251360.3EC78DF4@virgin.net>, > Alan Greig wrote: > >> I may actually be able to help here. Check your message composition settings > >> (I don't recall where they are, but the Netscape repferences are fairly easy > >> to deal with in both GUI and text forms) and tell it to always use plain text > >> rather than HTML. > > >Already set... > > Plan B: install a real operating system and use a real newsreader. > Nah just a reinstall of the existing one and apps usually fixes this type of problem with Windows. On another machine I read/post from wrapping is always fine but the spell-checker is jammed in US English and I want UK. I can change the setting dynamically but can't make it stick. Experience has shown me that these types of problems often boil down to some obscure registry problem. I used to try and track these things down but after 48 hours figuring out why Real Player settings didn't stick (two equal keys in the registry, one upper case, one lower - confused the hell out of windows) and similar with other problems I now follow the advice of our NT support guys. Either put up with the problem or re-install. The frustration of ploughing through undocumented registry settings isn't worth it. > > -- > Rev. Peter da Silva, ULC. "Cave cuniculos lagana ferentes" > > "Be conservative in what you generate, and liberal in what you accept" > -- Matthew 10:16 (l.trans) -- Alan Greig ###### From: peter@taronga.com (Peter da Silva) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 23 Dec 2001 11:43:00 GMT Organization: TSS Inc. Lines: 17 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <3C251360.3EC78DF4@virgin.net> <3C25947A.7FF3A951@virgin.net> NNTP-Posting-Host: citadel.in.taronga.com X-Trace: citadel.in.taronga.com 1009107780 18575 10.0.0.43 (23 Dec 2001 11:43:00 GMT) X-Complaints-To: usenet@taronga.com NNTP-Posting-Date: 23 Dec 2001 11:43:00 GMT X-Newsreader: trn 4.0-test72 (19 April 1999) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.stanford.edu!news-proxy.baileynm.com!citadel.in.taronga.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8650 In article <3C25947A.7FF3A951@virgin.net>, Alan Greig wrote: >Peter da Silva wrote: >> Plan B: install a real operating system and use a real newsreader. >Nah just a reinstall of the existing one and apps usually fixes this >type of problem with Windows. And there you have it. Would anyone have considered an OS where reinstalling the OS was the easiest way to fix a problem in a minor application, twenty years ago? How's that for progress. -- Rev. Peter da Silva, ULC. "Cave cuniculos lagana ferentes" "Be conservative in what you generate, and liberal in what you accept" -- Matthew 10:16 (l.trans) ###### Path: chonsp.franklin.ch!not-for-mail From: Neil Franklin Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 23 Dec 2001 18:30:56 +0100 Organization: My own Private Self Lines: 51 Message-ID: <6ug061zvvz.fsf@chonsp.franklin.ch> References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <9vsfqc$1ng$4@bob.news.rcn.net> <85bsgtosjc.fsf_-_@junk.nocrew.org> <9vvhri$k52$10@bob.news.rcn.net> <6u7kreaniz.fsf@chonsp.franklin.ch> <3C2442B2.20D5CE1E@jetnet.ab.ca> NNTP-Posting-Host: chonsp.franklin.ch X-Trace: chonsp.franklin.ch 1009128657 817 10.0.3.2 (23 Dec 2001 17:30:57 GMT) X-Complaints-To: news@chonsp.franklin.ch NNTP-Posting-Date: 23 Dec 2001 17:30:57 GMT X-Newsreader: Gnus v5.7/Emacs 20.4 Xref: chonsp.franklin.ch alt.sys.pdp10:8687 Ben Franchuk writes: > Neil Franklin wrote: > > > > Jan Vorbrueggen writes: > > > > > void any advantage an I-cache might bring. > > > > Why would that limit it? And why void the I-cache? > > > > > > I have programmed x86 and 86k, read a lot on ARM, MIPS, Sparc and PPC > > implementation, and am presently designing an PDP-10 microprocessor > > just for the fun of it. I see no reason why I can not achieve 1990 > > ARM/MIPS/Sparc style performance, from an 1990-chip-size-equvalent > > FPGA, using an 3-read-port register set. > > > Sounds fast! > Just how big is that FPGA to be 1990 equvalent? Well first I will define 1990 = 486DX-25 (I am using Intel for this calculation, as I have detailled data for them). For ARM/MIPS/Sparc aim for say same logic complexity but double speed. For 486DX-25 we have: 1.08mio transistors, which seem to be split roughly 64k*6=400k for the 8k*8bit SRAM cache, 200-300k for the FPU, rest 300-400k for the integer unit. So we need an FPGA with 64kbit BRAM and 150kGates (= 600k transistors when assuming standard 4-transistor NAND gates). And 25MHz for 486, 50MHz for ARM/MIPS/Sparc. The XC2S200 I am using[1] has 14*4k=56kbit BRAMs and 80kGates logic. So is between 1/2 the way and fully there. And I am programming it using low-level techniques, so I should get good logic utilisation. And I expect to get 50MHz from this FPGA. Now add that an PDP-10 is a lot simpler than an x86 (= uses less logic), but then subtract the space lost to putting chipset/peripherals into my FPGA, and I expect to get into 1990 speed range. [1] as crossposted to comp.os.vms and comp.sys.dec here the URL: http://neil.franklin.ch/Projects/PDP-10/ -- Neil Franklin, neil@franklin.ch.remove http://neil.franklin.ch/ Hacker, Unix Guru, El Eng HTL/BSc, Sysadmin, Archer, Roleplayer - Intellectual Property is Intellectual Robbery ###### Date: Wed, 26 Dec 2001 11:33:34 +0100 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Newsreader: Virtual Access by Atlantic Coast PLC, http://www.atlantic-coast.com/va Organization: Home Message-ID: Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) From: Paul Sture Reply-To: paul.sture@bluewin.ch References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <3C251360.3EC78DF4@virgin.net> <3C25947A.7FF3A951@virgin.net> NNTP-Posting-Host: 62.202.121.45 X-Trace: 26 Dec 2001 12:10:59 +0100, 62.202.121.45 Lines: 24 X-Complaints-To: abuse@bluewin.ch Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.bluewin.ch!awk.sture.ch!nobody Xref: chonsp.franklin.ch alt.sys.pdp10:8750 In article , Peter da Silva wrote: > In article <3C25947A.7FF3A951@virgin.net>, > Alan Greig wrote: > >Peter da Silva wrote: > >> Plan B: install a real operating system and use a real newsreader. > > >Nah just a reinstall of the existing one and apps usually fixes this > >type of problem with Windows. > > And there you have it. Would anyone have considered an OS where reinstalling > the OS was the easiest way to fix a problem in a minor application, twenty > years ago? How's that for progress. > Splutter, cough. Was recommended to do that by IBM more than 20 years ago when we had a mainframe printer problem. It was a hardware problem, but I still had to go through the hoop of reinstalling software to prove it wasn't software's fault. This was DOS/VSE, and they had the concept of "dirty copies" of programs. I don't recall the precise details, but seem to remember something about the paging model writing back data segments into the executable. ___ Paul Sture Switzerland ###### From: peter@taronga.com (Peter da Silva) Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) Date: 26 Dec 2001 14:05:18 GMT Organization: TSS Inc. Lines: 19 Message-ID: References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <3C25947A.7FF3A951@virgin.net> NNTP-Posting-Host: citadel.in.taronga.com X-Trace: citadel.in.taronga.com 1009375518 3125 10.0.0.43 (26 Dec 2001 14:05:18 GMT) X-Complaints-To: usenet@taronga.com NNTP-Posting-Date: 26 Dec 2001 14:05:18 GMT X-Newsreader: trn 4.0-test72 (19 April 1999) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.mailgate.org!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.stanford.edu!news-proxy.baileynm.com!citadel.in.taronga.com!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp10:8755 In article , Paul Sture wrote: >Splutter, cough. Was recommended to do that by IBM more than 20 years ago when >we had a mainframe printer problem. It was a hardware problem, but >I still had to go through the hoop of reinstalling software to prove >it wasn't software's fault. You had to reinstall the whole OS? Or just that one package? I can see reinstalling a program, that's easy and fast and doesn't impact other use of the machine (except on Windows, where Microsoft uses application vendors as OS update shippers, so you have to reboot). But reinstalling the OS? -- Rev. Peter da Silva, ULC. "Cave cuniculos lagana ferentes" "Be conservative in what you generate, and liberal in what you accept" -- Matthew 10:16 (l.trans) ###### Date: Sat, 29 Dec 2001 12:29:58 +0100 Newsgroups: comp.os.vms,alt.sys.pdp10,comp.sys.dec Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Newsreader: Virtual Access by Atlantic Coast PLC, http://www.atlantic-coast.com/va Organization: Home Message-ID: Subject: Re: PDP-10 architectural flaws? (was: VMS missing features) From: Paul Sture Reply-To: paul.sture@bluewin.ch References: <3roT7.207043$tf5.13617499@bin5.nnrp.aus1.giganews.com> <3C25947A.7FF3A951@virgin.net> NNTP-Posting-Host: 195.186.243.131 X-Trace: 29 Dec 2001 12:28:20 +0100, 195.186.243.131 Lines: 31 X-Complaints-To: abuse@bluewin.ch Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.bluewin.ch!awk.sture.ch!nobody Xref: chonsp.franklin.ch alt.sys.pdp10:8804 In article , Peter da Silva wrote: > In article , > Paul Sture wrote: > >Splutter, cough. Was recommended to do that by IBM more than 20 years ago when > >we had a mainframe printer problem. It was a hardware problem, but > >I still had to go through the hoop of reinstalling software to prove > >it wasn't software's fault. > > You had to reinstall the whole OS? Or just that one package? > Just stuff relevant to the printer, although there was probably more than one component. It was a political thing, as we knew that IBM were throwing their engineers at a major problem at the biggest IBM customer in town, and we were somewhat upset at the initial fobbing off. I did the reinstallation as a preemptive strike, so that on my next call I could simply say "Done that, please elevate the problem". IBM's elevation procedures worked beautifully :-) Given that it was indeed a h/w problem, the reinstallation was probably not needed, but it got us the required results. > I can see reinstalling a program, that's easy and fast and doesn't impact > other use of the machine (except on Windows, where Microsoft uses application > vendors as OS update shippers, so you have to reboot). But reinstalling the > OS? > No, definitely not the whole OS. That would have been time to grab the backup tapes. ___ Paul Sture Switzerland