Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Jupiter (was Re: Q: Why not (2^n)-bit?) References: <39A7357A.8E713045@ev1.net> <8o8hht$aau$1@lisa.gemair.com> <8o8ofo$i7o$1@pyrite.mv.net> <8o9797$t7e$1@lisa.gemair.com> <8o9oc0$ql2$1@pyrite.mv.net> <8odsdb$em$1@bob.news.rcn.net> <39ab70a9.86436118@news.newsguy.com> <8oisqr$9bs$1@bob.news.rcn.net> <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 31 Aug 2000 12:55:02 -0700 Message-ID: Lines: 8 X-Newsreader: Gnus v5.5/Emacs 20.3 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 31 Aug 2000 12:55:59 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.sys.pdp8:175 alt.sys.pdp10:357 alt.sys.pdp11:129 A.Greig@viirgin.net (Alan Greig) wrote: > Even the worst figures I've seen for the 2080 > put it at 10 times the performance of the VAX 11-780. jmfbahciv@aol.com writes: > Whoopy-skip. It wouldn't have run a TOPS operating system. Really? What would a 2080 have run, then? ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) Date: Sun, 10 Sep 00 07:56:06 GMT Organization: UltraNet Communications, Inc. Lines: 114 Message-ID: <8pfpmi$4s1$1@bob.news.rcn.net> References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> <39BA5626.BB6FD87C@bellatlantic.net> X-Trace: AMLxCaBUiMGrBHiSLMRsU1P9GsDUYmg3IFaO/k/1LX4= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 10 Sep 2000 11:01:38 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-spur1.maxwell.syr.edu!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!207-172-216-159 Xref: chonsp.franklin.ch alt.sys.pdp8:197 alt.sys.pdp10:401 alt.sys.pdp11:149 In article <39BA5626.BB6FD87C@bellatlantic.net>, hg/jb wrote: Bear with me. Hardware is not my game :-). And could you set up your replies with your answer at the bottom? I'm not going to yak very much if I have to expend most of my energy reformatting the damn post. >jmfbahciv@aol.com wrote: >> >> In article <8p9kpg$25cr$1@nntp1.ba.best.com>, >> inwap@best.com (Joe Smith) wrote: >> >In article , >> >Eric Smith wrote: >> >>A.Greig@viirgin.net (Alan Greig) wrote: >> >>> Even the worst figures I've seen for the 2080 >> >>> put it at 10 times the performance of the VAX 11-780. >> >> >> >>jmfbahciv@aol.com writes: >> >>> Whoopy-skip. It wouldn't have run a TOPS operating system. >> >> >> >>Really? What would a 2080 have run, then? >> > >> >The third-party 36-bit CPU from Systems Concepts was bug compatible with >> >a KL - you could take a unmodified SYSTEM.EXE and run it >> >on their hardware. >> >> Here's a question. Was the Jupiter 36-bits? I never saw one >> so I really don't know. >> >> > >> >The 2080 (and the XKL) were improvements on the KL design. >> >TOPS-10 had to be modified to run on the 1080 when the KL came out, and >> >modified again to run on the 2020 when the KS came out. >> >> I wouldn't call having to write a completely new CPU module >> as a modification, Joe :-). I usually thought of them as >> the CPU device driver; note that nobody else in my hearing >> ever called them that. Thus we had KASER, KISER, KLSER, and >> KSSER. >> >> >I think Barb's point is that even if the Jupiter hardware was ready, >> >the software was not. >> >> The device specific software could not be started until there >> was some hardware to run it on. It is a rule of computing that >> specs never matched the hardware. >> >> > Neither TOPS-10 or TOPS-20 would have run on the >> >2080 until after many man-months of tweaking the software. >> >> Tweaking, Joe? You might try rewriting. >> >> > And software >> >was not a priority (?!). >> >> Software was never a priority for the hardware types. We >> constantly had to fight to have a piece of new hardware >> put on the system so developing, and more importantly, >> testing could be done. There were always stories about >> delivery of a piece of gear to a customer, and astonishment >> on DEC's part when the customer asked for the software >> to be able to use it. >data path was 36bits. Which data path? >Address path was 30bits - I won't get into that. And what would the effective address be? >MBOX was (not counting ECC) 36bits to cache and 36bits to CPU but >quadword fetch from the memory array. I remember green app programmers wanting to arrange their code so that one would end up reading the code via a matrix. For example, say there are tasks A, B, C, D, each taking four instructions to complete. The page of code would read A1 B1 C1 D1 A2 B2 C2 D2 . . . C4 D4 Now just think of the AC problems with that. I was glad that Jupiter was cancelled if the architecture required that kind of MACRO-10 programming. >Cache was four way set associative. I remember this magic incantation. I have no idea what it means :-). Explanation? Oh, yeah. What kind of peripherals were going to hang off the Jupiter? They were going to have disks...weren't they? /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) Date: Sat, 09 Sep 00 10:57:10 GMT Organization: UltraNet Communications, Inc. Lines: 59 Message-ID: <8pdftp$jq1$5@bob.news.rcn.net> References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> X-Trace: 0xGp/Bml90nt/5rW4iBfrNIjV4K+Ow9PCW0KDHAZvnc= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 9 Sep 2000 14:02:33 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-spur1.maxwell.syr.edu!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!207-172-216-249 Xref: chonsp.franklin.ch alt.sys.pdp8:200 alt.sys.pdp10:404 alt.sys.pdp11:152 In article <8p9kpg$25cr$1@nntp1.ba.best.com>, inwap@best.com (Joe Smith) wrote: >In article , >Eric Smith wrote: >>A.Greig@viirgin.net (Alan Greig) wrote: >>> Even the worst figures I've seen for the 2080 >>> put it at 10 times the performance of the VAX 11-780. >> >>jmfbahciv@aol.com writes: >>> Whoopy-skip. It wouldn't have run a TOPS operating system. >> >>Really? What would a 2080 have run, then? > >The third-party 36-bit CPU from Systems Concepts was bug compatible with >a KL - you could take a unmodified SYSTEM.EXE and run it >on their hardware. Here's a question. Was the Jupiter 36-bits? I never saw one so I really don't know. > >The 2080 (and the XKL) were improvements on the KL design. >TOPS-10 had to be modified to run on the 1080 when the KL came out, and >modified again to run on the 2020 when the KS came out. I wouldn't call having to write a completely new CPU module as a modification, Joe :-). I usually thought of them as the CPU device driver; note that nobody else in my hearing ever called them that. Thus we had KASER, KISER, KLSER, and KSSER. >I think Barb's point is that even if the Jupiter hardware was ready, >the software was not. The device specific software could not be started until there was some hardware to run it on. It is a rule of computing that specs never matched the hardware. > Neither TOPS-10 or TOPS-20 would have run on the >2080 until after many man-months of tweaking the software. Tweaking, Joe? You might try rewriting. > And software >was not a priority (?!). Software was never a priority for the hardware types. We constantly had to fight to have a piece of new hardware put on the system so developing, and more importantly, testing could be done. There were always stories about delivery of a piece of gear to a customer, and astonishment on DEC's part when the customer asked for the software to be able to use it. /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) Date: Wed, 13 Sep 00 10:57:46 GMT Organization: UltraNet Communications, Inc. Lines: 40 Message-ID: <8po1fv$m7a$2@bob.news.rcn.net> References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> <39BA5626.BB6FD87C@bellatlantic.net> <39BD6B9A.22B21A35@bellatlantic.net> X-Trace: DuxteNqpQ1kftycA+4bqpEYSPlxpNP1RgyDmr3TydtA= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 13 Sep 2000 14:03:43 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-spur1.maxwell.syr.edu!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!207-172-216-202 Xref: chonsp.franklin.ch alt.sys.pdp8:208 alt.sys.pdp10:423 alt.sys.pdp11:160 In article <39BD6B9A.22B21A35@bellatlantic.net>, hg/jb wrote: >Eric Smith wrote: >> >> hg/jb writes: >> > Address path was 30bits - I won't get into that. >> >> If it supported 30-bit physical addresses, I'd very much like to >> know the details of the changes to the page table entries. >As Barb has said the software lagged the design, probably rather a >good thing too. And I guess you don't understand that the software _has_ to lag the hardware design. It not a good or bad thing; it's just a fact of life. >I do not recall the impact, I know we were focussing >on Fortran first. I know the Fortran people were getting talked to but that just a complete waste of time. Until you get an operating system working that the Fortran compiler and OTS and run under, the language people might just as well suck their thumbs. Instead some were coding for a ficticious Jupiter during their rewrite of FOROTS. This was a practice that I did manage to stop. I think the question had to do with how the addresses were laid out by the hardware. But I'm still munching on that explanation of 4-way associative cache so the probability that I'm wrong is 98%. I do know that the question had nothing to do with user apps, which is what Fortran is compared to an OS. /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) Date: Wed, 13 Sep 00 11:01:57 GMT Organization: UltraNet Communications, Inc. Lines: 33 Message-ID: <8po1nq$m7a$3@bob.news.rcn.net> References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> <39BA5626.BB6FD87C@bellatlantic.net> X-Trace: eJX5U3Fomo6lBjvLnVwMe/XqvIpK/kXcdi6yX8Id2Co= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 13 Sep 2000 14:07:54 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-spur1.maxwell.syr.edu!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!207-172-216-202 Xref: chonsp.franklin.ch alt.sys.pdp8:209 alt.sys.pdp10:424 alt.sys.pdp11:161 In article , Mark nospam@garetech.com.au wrote: >in article qh7l8ik0e8.fsf@ruckus.brouhaha.com, Eric Smith at >eric-no-spam-for-me@brouhaha.com wrote on 12/09/2000 08:51: > >> hg/jb writes: >>> Address path was 30bits - I won't get into that. >> >> If it supported 30-bit physical addresses, I'd very much like to >> know the details of the changes to the page table entries. >> >> Cheers, >> Eric > >Hi Eric, > vir address 30 bits is not a problem all the >pointers in memory had room >for 30 bits go look. At the phy lvl the core page >pointer only has room for >27 bit phy with out modifying the old format bits >18-22 in KL MBZ. The >problem is how TOPS now deals with 4096 section pointers in UPT/EPT. XKL >went for some method using "Super Section" pointers >in UPT/EPT and a second >lvl of indirection ( I think not an XKL guru). Then when did the effective address calculation happen? Now I wish I'd paid more attention to what Jim did when he messed with extended sections. /BAH Subtract a hundred and four for e-mail. ###### From: inwap@best.com (Joe Smith) Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) Date: 8 Sep 2000 03:01:04 GMT Organization: Chez Inwap Lines: 26 Message-ID: <8p9kpg$25cr$1@nntp1.ba.best.com> References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> NNTP-Posting-Host: shell3.ba.best.com X-Trace: nntp1.ba.best.com 968382064 71067 206.184.139.134 (8 Sep 2000 03:01:04 GMT) X-Complaints-To: abuse@best.com NNTP-Posting-Date: 8 Sep 2000 03:01:04 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!sunqbc.risq.qc.ca!headwall.stanford.edu!feeder.via.net!news1.best.com!nntp1.ba.best.com!inwap Xref: chonsp.franklin.ch alt.sys.pdp8:195 alt.sys.pdp10:389 alt.sys.pdp11:148 In article , Eric Smith wrote: >A.Greig@viirgin.net (Alan Greig) wrote: >> Even the worst figures I've seen for the 2080 >> put it at 10 times the performance of the VAX 11-780. > >jmfbahciv@aol.com writes: >> Whoopy-skip. It wouldn't have run a TOPS operating system. > >Really? What would a 2080 have run, then? The third-party 36-bit CPU from Systems Concepts was bug compatible with a KL - you could take a unmodified SYSTEM.EXE and run it on their hardware. The 2080 (and the XKL) were improvements on the KL design. TOPS-10 had to be modified to run on the 1080 when the KL came out, and modified again to run on the 2020 when the KS came out. I think Barb's point is that even if the Jupiter hardware was ready, the software was not. Neither TOPS-10 or TOPS-20 would have run on the 2080 until after many man-months of tweaking the software. And software was not a priority (?!). -Joe -- See http://www.inwap.com/ for PDP-10 and "ReBoot" pages. ###### Message-ID: <39BA5626.BB6FD87C@bellatlantic.net> From: hg/jb Reply-To: shsrms@bellatlantic.net Organization: The Keltic League X-Mailer: Mozilla 4.73 [en]C-CCK-MCD BA45DSL (WinNT; U) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 66 Date: Sat, 09 Sep 2000 15:23:56 GMT NNTP-Posting-Host: 138.88.34.164 X-Complaints-To: newsadmin@bellatlantic.net X-Trace: typhoon1.ba-dsg.net 968513036 138.88.34.164 (Sat, 09 Sep 2000 11:23:56 EDT) NNTP-Posting-Date: Sat, 09 Sep 2000 11:23:56 EDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!news-stu1.dfn.de!news-koe1.dfn.de!news.ruhrgebiet.individual.net!newsfeed.ision.net!ision!cpk-news-hub1.bbnplanet.com!washdc3-snf1!news.gtei.net!cyclone1.ba-dsg.net!typhoon1.ba-dsg.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp8:202 alt.sys.pdp10:406 alt.sys.pdp11:154 data path was 36bits. Address path was 30bits - I won't get into that. MBOX was (not counting ECC) 36bits to cache and 36bits to CPU but quadword fetch from the memory array. Cache was four way set associative. bo jmfbahciv@aol.com wrote: > > In article <8p9kpg$25cr$1@nntp1.ba.best.com>, > inwap@best.com (Joe Smith) wrote: > >In article , > >Eric Smith wrote: > >>A.Greig@viirgin.net (Alan Greig) wrote: > >>> Even the worst figures I've seen for the 2080 > >>> put it at 10 times the performance of the VAX 11-780. > >> > >>jmfbahciv@aol.com writes: > >>> Whoopy-skip. It wouldn't have run a TOPS operating system. > >> > >>Really? What would a 2080 have run, then? > > > >The third-party 36-bit CPU from Systems Concepts was bug compatible with > >a KL - you could take a unmodified SYSTEM.EXE and run it > >on their hardware. > > Here's a question. Was the Jupiter 36-bits? I never saw one > so I really don't know. > > > > >The 2080 (and the XKL) were improvements on the KL design. > >TOPS-10 had to be modified to run on the 1080 when the KL came out, and > >modified again to run on the 2020 when the KS came out. > > I wouldn't call having to write a completely new CPU module > as a modification, Joe :-). I usually thought of them as > the CPU device driver; note that nobody else in my hearing > ever called them that. Thus we had KASER, KISER, KLSER, and > KSSER. > > >I think Barb's point is that even if the Jupiter hardware was ready, > >the software was not. > > The device specific software could not be started until there > was some hardware to run it on. It is a rule of computing that > specs never matched the hardware. > > > Neither TOPS-10 or TOPS-20 would have run on the > >2080 until after many man-months of tweaking the software. > > Tweaking, Joe? You might try rewriting. > > > And software > >was not a priority (?!). > > Software was never a priority for the hardware types. We > constantly had to fight to have a piece of new hardware > put on the system so developing, and more importantly, > testing could be done. There were always stories about > delivery of a piece of gear to a customer, and astonishment > on DEC's part when the customer asked for the software > to be able to use it. > > /BAH > > Subtract a hundred and four for e-mail. ###### From: inwap@best.com (Joe Smith) Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) Date: 10 Sep 2000 01:44:56 GMT Organization: Chez Inwap Lines: 17 Message-ID: <8pep2o$2n70$1@nntp1.ba.best.com> References: <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> NNTP-Posting-Host: shell3.ba.best.com X-Trace: nntp1.ba.best.com 968550296 89312 206.184.139.134 (10 Sep 2000 01:44:56 GMT) X-Complaints-To: abuse@best.com NNTP-Posting-Date: 10 Sep 2000 01:44:56 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!sunqbc.risq.qc.ca!cpk-news-hub1.bbnplanet.com!news.gtei.net!feed1.news.rcn.net!rcn!news2.best.com!nntp1.ba.best.com!inwap Xref: chonsp.franklin.ch alt.sys.pdp8:198 alt.sys.pdp10:402 alt.sys.pdp11:150 In article <8pdftp$jq1$5@bob.news.rcn.net>, wrote: >Here's a question. Was the Jupiter 36-bits? I never saw one >so I really don't know. I never saw one either, but I heard a rumor that the schematic for the VAX 8600 CPU looked suspiciously like the schematic for the Jupiter with 4 bits disabled in the ALU. The schematics for the KS-2020 shows that it has a 40-bit ALU. Each half-word comprised five 4-bit bit-slice units. This had advantages; the extra high-order bits on the LH made it easier to correctly set CARRY0 and CARRY1 on an integer overflow condition and the extra low-order bits in the RH were useful when rounding floating point numbers. -Joe -- See http://www.inwap.com/ for PDP-10 and "ReBoot" pages. ###### Message-ID: <39BB9955.86133E71@bellatlantic.net> From: hg/jb Reply-To: shsrms@bellatlantic.net Organization: The Keltic League X-Mailer: Mozilla 4.73 [en]C-CCK-MCD BA45DSL (WinNT; U) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> <39BA5626.BB6FD87C@bellatlantic.net> <8pfpmi$4s1$1@bob.news.rcn.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 133 Date: Sun, 10 Sep 2000 14:22:48 GMT NNTP-Posting-Host: 138.88.34.164 X-Complaints-To: newsadmin@bellatlantic.net X-Trace: typhoon2.ba-dsg.net 968595768 138.88.34.164 (Sun, 10 Sep 2000 10:22:48 EDT) NNTP-Posting-Date: Sun, 10 Sep 2000 10:22:48 EDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!news.uni-stuttgart.de!uni-erlangen.de!newsfeeds.belnet.be!news.belnet.be!xfer13.netnews.com!netnews.com!newsfeed.skycache.com!Cidera!cyclone1.ba-dsg.net!typhoon2.ba-dsg.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp8:201 alt.sys.pdp10:405 alt.sys.pdp11:153 Ok, I been smacked, I will put my answers following your questions. jmfbahciv@aol.com wrote: > > In article <39BA5626.BB6FD87C@bellatlantic.net>, > hg/jb wrote: > > Bear with me. Hardware is not my game :-). And could you set > up your replies with your answer at the bottom? I'm not > going to yak very much if I have to expend most of my energy > reformatting the damn post. > > >jmfbahciv@aol.com wrote: > >> > >> In article <8p9kpg$25cr$1@nntp1.ba.best.com>, > >> inwap@best.com (Joe Smith) wrote: > >> >In article , > >> >Eric Smith wrote: > >> >>A.Greig@viirgin.net (Alan Greig) wrote: > >> >>> Even the worst figures I've seen for the 2080 > >> >>> put it at 10 times the performance of the VAX 11-780. > >> >> > >> >>jmfbahciv@aol.com writes: > >> >>> Whoopy-skip. It wouldn't have run a TOPS operating system. > >> >> > >> >>Really? What would a 2080 have run, then? > >> > > >> >The third-party 36-bit CPU from Systems Concepts was bug compatible > with > >> >a KL - you could take a unmodified SYSTEM.EXE and run it > >> >on their hardware. > >> > >> Here's a question. Was the Jupiter 36-bits? I never saw one > >> so I really don't know. > >> > >> > > >> >The 2080 (and the XKL) were improvements on the KL design. > >> >TOPS-10 had to be modified to run on the 1080 when the KL came out, and > >> >modified again to run on the 2020 when the KS came out. > >> > >> I wouldn't call having to write a completely new CPU module > >> as a modification, Joe :-). I usually thought of them as > >> the CPU device driver; note that nobody else in my hearing > >> ever called them that. Thus we had KASER, KISER, KLSER, and > >> KSSER. > >> > >> >I think Barb's point is that even if the Jupiter hardware was ready, > >> >the software was not. > >> > >> The device specific software could not be started until there > >> was some hardware to run it on. It is a rule of computing that > >> specs never matched the hardware. > >> > >> > Neither TOPS-10 or TOPS-20 would have run on the > >> >2080 until after many man-months of tweaking the software. > >> > >> Tweaking, Joe? You might try rewriting. > >> > >> > And software > >> >was not a priority (?!). > >> > >> Software was never a priority for the hardware types. We > >> constantly had to fight to have a piece of new hardware > >> put on the system so developing, and more importantly, > >> testing could be done. There were always stories about > >> delivery of a piece of gear to a customer, and astonishment > >> on DEC's part when the customer asked for the software > >> to be able to use it. > > >data path was 36bits. > > Which data path? Sorry, the effective, ahem I have to say this I know I will get smacked again... the effectively emulated KC10 datapath. > > >Address path was 30bits - I won't get into that. > > And what would the effective address be? I don't know the answer to this one. > > >MBOX was (not counting ECC) 36bits to cache and 36bits to CPU but > >quadword fetch from the memory array. > > I remember green app programmers wanting to arrange their > code so that one would end up reading the code via > a matrix. For example, say there are tasks A, B, C, D, each > taking four instructions to complete. The page of code > would read > > A1 > B1 > C1 > D1 > A2 > B2 > C2 > D2 > . > . > . > C4 > D4 > > Now just think of the AC problems with that. I was glad that > Jupiter was cancelled if the architecture required that kind > of MACRO-10 programming. And that is what bit the design too apparently. I had thought it was just keeping the pipe full but there was more to the story and branch handling was a bit less than perfect. > > >Cache was four way set associative. > > I remember this magic incantation. I have no idea what it means :-). > Explanation? I will have to go back to my books to dig out the english explanation and hope someone can chime in here. > > Oh, yeah. What kind of peripherals were going to hang off > the Jupiter? They were going to have disks...weren't they? Yes, there was an I/O box that was essentially a converter from a TTL bus to 100K ECL bus, with a bit of an elastic store. I used a first in first out (FIFO hardware) design to accomodate speed mismatch between unibus, mass bus, CI, and we hoped - IBM channel (like DX11) interfaces. Hanging off this bus were unibus, massbus, ci adapters that allowed the connection of off the shelf controllers. HSC was considered to be the storage beast of choice by corp with CI connectivity to it. bob> > /BAH > > Subtract a hundred and four for e-mail. ###### From: rivie@server.newlogan.teraglobal (Roger Ivie) Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> <39BA5626.BB6FD87C@bellatlantic.net> <8pfpmi$4s1$1@bob.news.rcn.net> Reply-To: rivie@teraglobal.com Message-ID: User-Agent: slrn/0.9.6.2 (FreeBSD) NNTP-Posting-Host: 208.186.13.23 Date: 11 Sep 2000 10:38:15 -0500 X-Trace: 11 Sep 2000 10:38:15 -0500, 208.186.13.23 Lines: 30 X-Comments: This message was posted through Newsfeeds.com X-Comments2: IMPORTANT: Newsfeeds.com does not condone, nor support, spam or any illegal or copyrighted postings. X-Comments3: IMPORTANT: Under NO circumstances will postings containing illegal or copyrighted material through this service be tolerated!! X-Report: Please report illegal or inappropriate use to X-Abuse-Info: Please be sure to forward a copy of ALL headers, INCLUDING the body (DO NOT SEND ATTACHMENTS) Organization: Newsfeeds.com http://www.newsfeeds.com 80,000+ UNCENSORED Newsgroups. Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!uni-erlangen.de!newsfeed.germany.net!newsfeed.wirehub.nl!fr.usenet-edu.net!usenet-edu.net!local-out.newsfeeds.com!newsfeeds.com!corp.newsfeeds.com!newsfeeds.com!crp!anonymous!127.0.0.1!crp!rivie Xref: chonsp.franklin.ch alt.sys.pdp8:204 alt.sys.pdp10:416 alt.sys.pdp11:156 In article <8pfpmi$4s1$1@bob.news.rcn.net>, jmfbahciv@aol.com wrote: >In article <39BA5626.BB6FD87C@bellatlantic.net>, > hg/jb wrote: >>Cache was four way set associative. > >I remember this magic incantation. I have no idea what it means :-). >Explanation? This means that any given main memory address may be present in any of four lines of the cache. In a "normal" (one-way set associative?) cache, each memory address may be in one and only one place in the cache. So if you access a memory location that happens to be mapped by the same location in the cache as something else you're interested in, that something else has to be tossed out to make room for it. With a four-way set associative cache, that doesn't have to happen (unless you're touching five things that all happen to map to the same cache location...). -- Roger Ivie TeraGlobal Communications Corporation 1770 North Research Park Way Suite 100 Logan, UT 84341 mailto:rivie@teraglobal.com phoneto:(435)787-0555 faxto:(435)787-0516 -----= Posted via Newsfeeds.Com, Uncensored Usenet News =----- http://www.newsfeeds.com - The #1 Newsgroup Service in the World! -----== Over 80,000 Newsgroups - 16 Different Servers! =----- ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> <39BA5626.BB6FD87C@bellatlantic.net> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 11 Sep 2000 15:51:43 -0700 Message-ID: Lines: 8 X-Newsreader: Gnus v5.5/Emacs 20.3 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 11 Sep 2000 15:50:27 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!nntp-out.monmouth.com!newspeer.monmouth.com!newsfeed.direct.ca!look.ca!news-peer-west.sprintlink.net!news.sprintlink.net!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.sys.pdp8:207 alt.sys.pdp10:421 alt.sys.pdp11:159 hg/jb writes: > Address path was 30bits - I won't get into that. If it supported 30-bit physical addresses, I'd very much like to know the details of the changes to the page table entries. Cheers, Eric ###### Message-ID: <39BD6B9A.22B21A35@bellatlantic.net> From: hg/jb Reply-To: shsrms@bellatlantic.net Organization: The Keltic League X-Mailer: Mozilla 4.73 [en]C-CCK-MCD BA45DSL (WinNT; U) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> <39BA5626.BB6FD87C@bellatlantic.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 15 Date: Mon, 11 Sep 2000 23:32:05 GMT NNTP-Posting-Host: 138.88.35.169 X-Complaints-To: newsadmin@bellatlantic.net X-Trace: typhoon2.ba-dsg.net 968715125 138.88.35.169 (Mon, 11 Sep 2000 19:32:05 EDT) NNTP-Posting-Date: Mon, 11 Sep 2000 19:32:05 EDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!uni-erlangen.de!newsfeeds.belnet.be!news.belnet.be!xfer13.netnews.com!netnews.com!newsfeed.skycache.com!Cidera!cyclone2.ba-dsg.net!typhoon2.ba-dsg.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp8:205 alt.sys.pdp10:417 alt.sys.pdp11:157 Eric Smith wrote: > > hg/jb writes: > > Address path was 30bits - I won't get into that. > > If it supported 30-bit physical addresses, I'd very much like to > know the details of the changes to the page table entries. As Barb has said the software lagged the design, probably rather a good thing too. I do not recall the impact, I know we were focussing on Fortran first. bob > > Cheers, > Eric ###### From: Mark nospam@garetech.com.au Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) Organization: Garetech Computer Solutions Lines: 24 Message-ID: References: <39ae2fab.266407563@news.newsguy.com> <8olg3a$8f8$1@bob.news.rcn.net> <8p9kpg$25cr$1@nntp1.ba.best.com> <8pdftp$jq1$5@bob.news.rcn.net> <39BA5626.BB6FD87C@bellatlantic.net> Mime-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit User-Agent: Microsoft-Outlook-Express-Macintosh-Edition/5.02.2022 Date: Mon, 11 Sep 2000 23:37:47 GMT NNTP-Posting-Host: 144.132.247.196 X-Complaints-To: news@bigpond.net.au X-Trace: news-server.bigpond.net.au 968715467 144.132.247.196 (Tue, 12 Sep 2000 10:37:47 EST) NNTP-Posting-Date: Tue, 12 Sep 2000 10:37:47 EST Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!news.tele.dk!139.130.240.98!intgwlon.nntp.telstra.net!news-server.bigpond.net.au!not-for-mail Xref: chonsp.franklin.ch alt.sys.pdp8:203 alt.sys.pdp10:415 alt.sys.pdp11:155 in article qh7l8ik0e8.fsf@ruckus.brouhaha.com, Eric Smith at eric-no-spam-for-me@brouhaha.com wrote on 12/09/2000 08:51: > hg/jb writes: >> Address path was 30bits - I won't get into that. > > If it supported 30-bit physical addresses, I'd very much like to > know the details of the changes to the page table entries. > > Cheers, > Eric Hi Eric, vir address 30 bits is not a problem all the pointers in memory had room for 30 bits go look. At the phy lvl the core page pointer only has room for 27 bit phy with out modifying the old format bits 18-22 in KL MBZ. The problem is how TOPS now deals with 4096 section pointers in UPT/EPT. XKL went for some method using "Super Section" pointers in UPT/EPT and a second lvl of indirection ( I think not an XKL guru). Cheers Mark :) ###### From: inwap@best.com (Joe Smith) Newsgroups: comp.os.vms,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Jupiter (was Re: Q: Why not (2^n)-bit?) Date: 14 Sep 2000 07:31:42 GMT Organization: Chez Inwap Lines: 56 Message-ID: <8ppusu$1id6$1@nntp1.ba.best.com> References: , wrote: >In article , > Mark nospam@garetech.com.au wrote: >> vir address 30 bits is not a problem all the pointers in memory >>had room for 30 bits go look. At the phy lvl the core page pointer >>only has room for 27 bit phy with out modifying the old format bits >>18-22 in KL MBZ. The problem is how TOPS now deals with 4096 section >>pointers in UPT/EPT. XKL went for some method using "Super Section" >>pointers in UPT/EPT and a second lvl of indirection. > >Then when did the effective address calculation happen? >Now I wish I'd paid more attention to what Jim did when >he messed with extended sections. Extended sections did not change when the EA calculation was done. Here's a quote from the Processor Reference Manual: PLEASE READ THIS The calculation of E is the first step in the execution of every instruction. No other action taken by any instruction, no matter what it is, can possibly precede that calculation. There is absolutely nothing whatsover that any instruction can do to any accumulator or memory location that can in any way affect its own effective address calculation. In particular, will add one to accumulator T1, then jump to the effective address that was previously calculated using the former contents of T1. It also means that the 30-bit effective address calculation occurs before instruction decode starts. (Section zero or not.) Remember, the KL has a 30-bit virtual address space. A 30-bit address is calculated by using 1) the current PC with an 18-bit offset, 2) a 35-bit (positive) index register, or 3) by using a 1+1+4+30 bit global indirect word. The 30 bits are a 18-bit offset and a 12-bit section number. But only sections 0-37 and 7777 have any meaning. If I understand page 1-27 correctly, then a 30-bit address of 41,,777777 was treated the same as an address of 1,,777777 - but any reference to section 7777 would get a page fault, regardless of whether section 37 was mapped or not. As Mark@garetech wrote, the main question is "how does the CPU use the UPT/EPT to translate a 30-bit virtual address to a physical address?" XKL uses bits 6-8 to select one of 8 super-section pointers from the UPT/EPT, then bits 9-17 to select one of 512 sections from a super-section page map, then the usual 9 bits of page within section and 9 bits of word within page. -Joe -- See http://www.inwap.com/ for PDP-10 and "ReBoot" pages.