From: pm215@cam.ac.uk (Peter Maydell) Newsgroups: alt.folklore.computers Subject: word sizes (was: Re: Why IBM chose Intel) Date: 24 Jan 1998 02:29:45 -0000 Organization: Cambridge University Lines: 36 Message-ID: <6abjmq$6j8@mnementh.trin.cam.ac.uk> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> NNTP-Posting-Host: mnementh.trin.cam.ac.uk Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!btnet-peer!btnet!server2.netnews.ja.net!lyra.csx.cam.ac.uk!not-for-mail In article <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk>, wrote: >On 1998-01-23 Neil.Franklin.remove.this@ccw.ch said: > -Dorado had up to 16Mbyte momory, that suggests 32 bit (or at least > -24bit) architecture. Dolphin had "more momory that Alto". But these > -ware all non-uP systems. Dandelion (Xerox Star 8010) still was made > -from 2901 bit slices. > >that's the danger of trying to guess word length supported from memory >space available. the dorado was basically what they dreamed the alto >would become; it did stick the the alto's basic architecture in some >ways, and the accumulators (actually 256-deep ac stacks) were 16 bits >wide. also, dorado smalltalk-80 used 15-bit addresses (well, table >entries) and smallints with a flag to discriminate between them. Indeed, word lengths are funny things. Take the PERQ, for example. CPU registers were 20 bits wide. Memory was 64 bits wide, accessed over a 16bit bus (you could transfer single, double or quad words; quad words were most efficient as you only had one lot of setup time. The rasterop hardware performs some funky overlapped accesses to allow it to read source and destination quadwords, combine them and write the result back to the destination, in 12 microcycles.) The microcode implementation could choose any word size it wanted (there were 256 hardware registers so you could easily afford to implement a 32bit word in two 20bit registers.) I think 16 bits was what was actually used for POS and unix. The address bus to memory was 20 bits, I think. The microcode address bus was 14 bits wide (12 in earlier versions of the CPU board) and the microcode word was 48 bits wide. And since the microcode was designed for bytecode interpretation the actual opcodes would be 8 bits wide. And then there was a prototype system which extended all the 20bit stuff to 24 bits... Peter Maydell [I seem to be mentioning the PERQ every time I post here; it really is an elegant piece of hardware, though. And the graphics performance is amazing.] ###### From: spam@lisardrock.demon.co.uk Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Why IBM chose Intel) Date: Sat, 24 Jan 1998 15:13:46 GMT Message-ID: <885654826.29133.2.nnrp-09.9e98ee68@news.demon.co.uk> References: <6abjmq$6j8@mnementh.trin.cam.ac.uk> NNTP-Posting-Host: lisardrock.demon.co.uk X-NNTP-Posting-Host: lisardrock.demon.co.uk [158.152.238.104] X-Everything: Net-Tamer V 1.09.2 Lines: 33 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!leto.ou.edu!news.onenet.net!news.oru.edu!dispose.news.demon.net!demon!news.demon.co.uk!demon!lisardrock.demon.co.uk!not-for-mail On 1998-01-24 pm215@cam.ac.uk(PeterMaydell) said: -Indeed, word lengths are funny things. Take the PERQ, for example. YES PLEASE! we've been nagging for a while now for one of those :> ah, but you already know this... -CPU registers were 20 bits wide. Memory was 64 bits wide, accessed -over a 16bit bus (you could transfer single, double or quad words; -quad words were most efficient as you only had one lot of setup -time. The rasterop hardware performs some funky overlapped accesses -to allow it to read source and destination quadwords, combine them -and write the result back to the destination, in 12 microcycles.) -The microcode implementation could choose any word size it wanted -(there were 256 hardware registers so you could easily afford to -implement a 32bit word in two 20bit registers.) I think 16 bits was -what was actually used for POS and unix. The address bus to memory -was 20 bits, I think. The microcode address bus was 14 bits wide -(12 in earlier versions of the CPU board) and the microcode word -was 48 bits wide. And since the microcode was designed for bytecode -interpretation the actual opcodes would be 8 bits wide. it does sound like a truly beautiful architecture. obviously the only way to convince oneself of that is to write something for it... anybody have a copy of the perq faqs within easy reach of their mailer? preferably the one with details of how you program the microcode... -- Communa - all at lisardrock.demon.net to be sure we read your reply, send to username `communa' Net-Tamer V 1.09.2 - Test Drive ###### From: ard@p850ug1.demon.co.uk (Tony Duell) Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: 24 Jan 1998 22:41:25 -0000 Organization: P850 User Group Message-ID: <6adqml$r2@p850ug1.demon.co.uk> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> NNTP-Posting-Host: p850ug1.demon.co.uk X-NNTP-Posting-Host: p850ug1.demon.co.uk [158.152.97.199] X-Newsreader: TIN [version 1.2 PL2] Lines: 65 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!dispose.news.demon.net!demon!news.demon.co.uk!demon!p850ug1.demon.co.uk!not-for-mail : Indeed, word lengths are funny things. Take the PERQ, for example. : CPU registers were 20 bits wide. Memory was 64 bits wide, accessed over It's a bit more complex than that - the CPU registers and ALU are 20 bits wide, but some parts of the CPU - the barrel shifter, the path to the main memory, etc are only 16 bits wide. Some of the flags (carry, etc) are updated based on the 16 bit result only. : a 16bit bus (you could transfer single, double or quad words; quad words There are 4 types of memory cycle - single, double, quad and quad reversed. The last one is the same as a quadword access except that the words are passed in the reverse order. The rasterop hardware uses quadword accesses, and it needs the reverse-order cycles when you do a rasterop right-to-left (I think) - which is needed sometimes when the source and destination regions overlap. : any word size it wanted (there were 256 hardware registers so you could : easily afford to implement a 32bit word in two 20bit registers.) I think Which would make life easier than say splitting it 20/12, as transfering a 20 bit register to the 16 bit memory is a minor kludge. : 16 bits was what was actually used for POS and unix. The address bus to : memory was 20 bits, I think. The microcode address bus was 14 bits wide Yes, 20 bit addresses (which is why the CPU was 20 bit), addressing by 16 bit words. So a maximum memory of 2Mbytes. There was no separate MMU - you emulated one using the CPU microcode and the normal ALU etc to relocate addresses. In fact the main reason for designing the 24 bit machine was to allow a larger address bus - for initially a 4Mbyte machine, and later (I believe) a 32Mbyte one. : (12 in earlier versions of the CPU board) and the microcode word was 48 : bits wide. And since the microcode was designed for bytecode interpretation : the actual opcodes would be 8 bits wide. There was hardware support for bytecode interpretation - there was a 8 byte store in the CPU that was loaded as 4 16 bit words (yes, a quadword access again) and read out as 8 bytes. These went into some logic and then into the J-mux (Jump Multiplexer), which allowed a 256-way microcode dispatch in one instruction. : And then there was a prototype system which extended all the 20bit stuff : to 24 bits... But, AFAIK, the 16 bit parts remained at 16 bits : Peter Maydell : [I seem to be mentioning the PERQ every time I post here; it really is an : elegant piece of hardware, though. And the graphics performance is amazing.] Well, the older 'classic' PERQs (which is what we're talking about here) are certainly fine machines. I'm not convinced that the PERQ AGW3300 (aka PERQ 3a) was so nice - it was a 68020-based unix box. Although it did have a _strange_ rasterop machine with a pair of 29116 processors, one for calculating addresses, the other for doing the rasterop. -tony ###### Path: ccw.ch!usenet From: Neil.Franklin.remove.this@ccw.ch Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Why IBM chose Intel) Date: 25 Jan 1998 20:23:13 +0100 Organization: My own Private Self Lines: 39 Message-ID: References: <6abjmq$6j8@mnementh.trin.cam.ac.uk> <885654826.29133.2.nnrp-09.9e98ee68@news.demon.co.uk> X-Newsreader: Gnus v5.3/Emacs 19.34 On 1998-01-24 pm215@cam.ac.uk(PeterMaydell) said: -Indeed, word lengths are funny things. Take the PERQ, for example. spam@lisardrock.demon.co.uk became excited: > YES PLEASE! we've been nagging for a while now for one of those :> ah, > but you already know this... This triggered a few thoughts (allways a bad thing): In 1991 there was a firm called Algotronix in Edinburgh. They made a form of SRAM controlled programmable gate array called CAL1024 (according to Byte Sept 91, page 88IS55, IS = International Supplement, titled: Algotronix, the first custom computer). These consisted of an 32x32 grid of cells with each: 4 input lines (their 4 neighbors), an 8 bit control (2 bit selects input A from 0..3, 2 bit for input B, 4 bit function (and/or/xor/nand/d-ff/...) and 1 output (sent to all neighbors). Input to and output of edge-of-grid cells via IO pins. One chips was claimed to be equivalent to 5000 gates or 20000 transistors of custom chip. They actually sold a board with 4x2 of these and 2 MB SRAM connected to some of the pins. The whole thing was an PC slot card (ISA). Think about how big one could make such an grid of cells with todays 10mio transistor technology. Then get detailed architecture descriptions of all the interesting things pdp6, pdp8, pdp10, pdp11, Alto, PERQ, 1401, 360 or anything else an a.f.c-er would like and make control maps for them. Then distribute the maps via Web together with original microcode and OSes. Neil "dreaming and getting frustrated" Franklin -- Neil.Franklin.remove.this@ccw.ch, http://www.ccw.ch/Neil.Franklin/ for Geek Code, Papernet, Voicenet, PGP public key see http: Mac, 95 and NT users are CLUEless (Command Line User Environment) If I go missing, its once again my newsfeed that has craped ###### From: newcomer@flounder.com (Joseph M. Newcomer) Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Mon, 26 Jan 1998 04:53:56 GMT Organization: Pittsburgh OnLine, Inc. Lines: 9 Message-ID: <34fb169d.464331033@206.210.64.12> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> NNTP-Posting-Host: ppp46.s8.pgh.net Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/32.451 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!howland.erols.net!fastnet!news.pgh.net!not-for-mail OK, here's a question that will demonstrate if you are a newbie or a genuine OF: Why did so many machines, e.g., the 709, 7090, PDP-6, PDP-10, and DEC-20 all have 36-bit words... joe Joseph M. Newcomer newcomer@flounder.com http://www3.pgh.net/~newcomer ###### From: newcomer@flounder.com (Joseph M. Newcomer) Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Mon, 26 Jan 1998 04:57:47 GMT Organization: Pittsburgh OnLine, Inc. Lines: 11 Message-ID: <34fc172e.464475729@206.210.64.12> References: <6abjmq$6j8@mnementh.trin.cam.ac.uk> <19980125052600.AAA27893@ladder02.news.aol.com> NNTP-Posting-Host: ppp46.s8.pgh.net Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Newsreader: Forte Agent 1.5/32.451 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!newsxfer.visi.net!feed2.news.erols.com!howland.erols.net!fastnet!news.pgh.net!not-for-mail Tom McWilliams, designer of the S-1 multiprocessor, once said to me "I made a serious mistake. I designed this machine for a 32-bit address space. But at today's prices, that could be populated for only $10,000,000, which is nothing to DOE". This was in the late 1970s, if I recall correctly. Figure it out: at today's prices, what does 4GB cost? What will it cost in two years? Joseph M. Newcomer newcomer@flounder.com http://www3.pgh.net/~newcomer ###### From: "George R. Gonzalez" Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Mon, 26 Jan 1998 09:34:50 -0600 Organization: SkyPoint Communications, Inc. Lines: 32 Message-ID: <6aiadv$t2g$1@shadow.skypoint.net> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> NNTP-Posting-Host: dial104.skypoint.net Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Newsreader: Microsoft Outlook Express 4.71.1712.3 X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!nntp.flash.net!news.he.net!news.pagesat.net!skypoint.com!not-for-mail Joseph M. Newcomer wrote in message <34fb169d.464331033@206.210.64.12>... >OK, here's a question that will demonstrate if you are a newbie or a >genuine OF: > >Why did so many machines, e.g., the 709, 7090, PDP-6, PDP-10, and >DEC-20 all have 36-bit words... > joe >Joseph M. Newcomer A multiple of the character size is convenient. 6-bit character sets were common back then, so you'd like a word size of 6,12,18,24,30,36,42,48,54, or 60. You also would like to get at least one address and one op-code in a word if possible. 64 op-codes would be nice, so that's 6 bits; 32K of addressing would be nice, so that's 15 more bits.. so you'd like to have at least 21 bits, maybe a few more for specifying index registers, so that narrows it down to 24,30,36,42,48,54, or 60. I can't think offhand of any 30, 42, or 54 bit computers, but there probably were some. That leaves the more common 24, 36, 48, and 60 bit sizes. Am I missing something? ###### From: "George R. Gonzalez" Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Mon, 26 Jan 1998 09:38:59 -0600 Organization: SkyPoint Communications, Inc. Lines: 21 Message-ID: <6aiall$t73$1@shadow.skypoint.net> References: <6abjmq$6j8@mnementh.trin.cam.ac.uk> <19980125052600.AAA27893@ladder02.news.aol.com> <34fc172e.464475729@206.210.64.12> NNTP-Posting-Host: dial104.skypoint.net Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Newsreader: Microsoft Outlook Express 4.71.1712.3 X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news.he.net!news.pagesat.net!skypoint.com!not-for-mail Joseph M. Newcomer wrote in message <34fc172e.464475729@206.210.64.12>... >Tom McWilliams, designer of the S-1 multiprocessor, once said to me "I >made a serious mistake. I designed this machine for a 32-bit address >space. But at today's prices, that could be populated for only >$10,000,000, which is nothing to DOE". This was in the late 1970s, if >I recall correctly. > >Figure it out: at today's prices, what does 4GB cost? What will it >cost in two years? CompUSA was giving away 16MB EDO SIMMS a while back. Get 256 of your friends to show up and you have your 4GB for $0.00 The first memory I bought was 1K bit 2102 static RAM, back in 1975 or so. It was about $10 a chip then and considered a real bargain. Quite a drop since then! ###### From: skb@xmission.removethis.com (Scott Brown) Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Mon, 26 Jan 1998 10:29:01 GMT Organization: XMission Internet (801 539 0900) Lines: 44 Message-ID: <34cc5b87.12617520@news.xmission.com> References: <6abjmq$6j8@mnementh.trin.cam.ac.uk> <19980125052600.AAA27893@ladder02.news.aol.com> <34fc172e.464475729@206.210.64.12> NNTP-Posting-Host: slc328.modem.xmission.com X-Newsreader: Forte Free Agent 1.1/16.230 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!newsxfer.visi.net!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!newsfeed.internetmci.com!207.0.56.122!news.eli.net!inquo!xmission!not-for-mail On Mon, 26 Jan 1998 04:57:47 GMT, newcomer@flounder.com (Joseph M. Newcomer) wrote: >Tom McWilliams, designer of the S-1 multiprocessor, once said to me "I >made a serious mistake. I designed this machine for a 32-bit address >space. But at today's prices, that could be populated for only >$10,000,000, which is nothing to DOE". This was in the late 1970s, if >I recall correctly. > >Figure it out: at today's prices, what does 4GB cost? What will it >cost in two years? A local supplier sells 8Mb, 72-pin DRAM SIMMs for $13 each, so four gigs of them cost only $6,656. That, of course, doesn't count the interface circuitry, power supply, and cooling (which is definitely non-trivial for 512 SIMMs). Not to mention physical real estate; those things are bulky. The same supplier sells 128Mb SDRAM DIMMs for $357 each, making for a 4Gb array costing $11,424 and comprised of just 32 DIMMs. That's downright practical. As for the future...who knows? It was only about a year ago (IIRC, which is not guaranteed) that I paid $100 for a 4Mb SIMM and considered it a good deal. My dealer now sells the same SIMM for eight bucks - and I've got about fifty megs' worth of them in my parts bin because I've got noplace to use them. If memory prices drop like that, in two years our 4Gb array will cost under a hundred dollars, and it will probably be an array of two credit-card-sized modules. I really don't know if memory prices will continue to drop so steeply. I keep thinking that we've hit this plateau where software demand for memory has kind of levelled off, so that adding more memory doesn't neccesarily gain you more performance. I just have to remind myself about how many times I've thought that before...like when I got from 4 to 16Mb of main memory, or when I went from a 40 to a 540Mb hard drive, or my 286->486 upgrade. All of those once-new parts were displaced in turn (by 64Mb, 3.5Gb, a K6-200) and are now part of my "old, slow" machine that runs FreeBSD. I wonder how long it will be before my fine new K6-200 is old and slow. -Scott ###### From: John Ahlstrom Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Tue, 27 Jan 1998 13:35:13 -0800 Organization: Cisco Systems, Inc. Lines: 49 Message-ID: <34CE5311.15FB7483@cisco.com> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aku43$bl0@dfw-ixnews10.ix.netcom.com> NNTP-Posting-Host: waterboy.cisco.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 3.0 (X11; I; SunOS 4.1.4 sun4m) To: John C Green Jr Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!leto.ou.edu!hammer.uoregon.edu!arclight.uoregon.edu!newsfeed.internetmci.com!4.1.16.34!cpk-news-hub1.bbnplanet.com!su-news-feed1.bbnplanet.com!news.bbnplanet.com!news-master.cisco.com!not-for-mail All: 1. Was the 704 the first 36-bit word computer? If not, what was? 2. Did the 6-bit char preceed the first 36-bit computer whatever it was? 3. Does anyone have a reference about the design of the first 36-bit computer that talks about 6-bit chars? Thanks John Ahlstrom jahlstrom@cisco.com > Joe, > That's only two architectures. There indeed were lots of 36-bit > machines: > > 704/709/7040/7044/7090/7094/... > PDP-6/KA10/KI10/KL10/... > GE-635/645/... > Univac 1106/1107/1108/1110/... > > and many others. > > As will be stated in several other posts it's because of 6-bit > characters (upper case only, ASR-33 style). 18-bit address > (256K words), a few bits for opcode, a few bits for index > register, storing two addresses (CAR/CDR) in one word, made > 36-bits a pretty logical length. > > -- > John C Green Jr > Internet Marketing and Business Development > Website Development and Project Management > 21483 Old Mine Rd (408)353-1870 > Los Gatos CA 95030 JCGreen@ix.netcom.com -- John Ahlstrom jahlstrom@cisco.com 408-526-6025 Using Java to Decrease Entropy As Moses knew, you can accomplish almost anything, if you don't care who gets the credit. Old Swedish Proverb ###### From: JCGreen@ix.netcom.com (John C Green Jr) Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Tue, 27 Jan 1998 15:18:43 GMT Organization: Netcom Lines: 37 Message-ID: <6aku43$bl0@dfw-ixnews10.ix.netcom.com> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> NNTP-Posting-Host: sjx-ca65-28.ix.netcom.com X-NETCOM-Date: Tue Jan 27 9:22:43 AM CST 1998 X-Newsreader: Forte Free Agent 1.0.82 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news-dc-3.sprintlink.net!news-dc-1.sprintlink.net!news-east.sprintlink.net!news-peer.sprintlink.net!news.sprintlink.net!Sprint!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!ix.netcom.com!news newcomer@flounder.com (Joseph M. Newcomer) wrote: >OK, here's a question that will demonstrate if you are a newbie or a >genuine OF: >Why did so many machines, e.g., the 709, 7090, PDP-6, PDP-10, and >DEC-20 all have 36-bit words... Joe, That's only two architectures. There indeed were lots of 36-bit machines: 704/709/7040/7044/7090/7094/... PDP-6/KA10/KI10/KL10/... GE-635/645/... Univac 1106/1107/1108/1110/... and many others. As will be stated in several other posts it's because of 6-bit characters (upper case only, ASR-33 style). 18-bit address (256K words), a few bits for opcode, a few bits for index register, storing two addresses (CAR/CDR) in one word, made 36-bits a pretty logical length. The PC generation believes God created the 8-bit byte on about the third day of Genesis and is unaware of a couple of decades of 6-bit characters and 12-, 18-, 24-, 36-, 48-, and 60-bit words that preceeded the almost universal adoption of 8-bit bytes and 16-, 32- and 64-bit word lengths. -- John C Green Jr Internet Marketing and Business Development Website Development and Project Management 21483 Old Mine Rd (408)353-1870 Los Gatos CA 95030 JCGreen@ix.netcom.com ###### Path: ccw.ch!usenet From: Neil.Franklin.remove.this@ccw.ch Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: 27 Jan 1998 23:21:18 +0100 Organization: My own Private Self Lines: 29 Message-ID: <4t2py2w1.fsf@chonsp.franklin.lugs.ch> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aiadv$t2g$1@shadow.skypoint.net> X-Newsreader: Gnus v5.3/Emacs 19.34 "George R. Gonzalez" asked: >I can't think offhand of any 30, 42, or 54 bit computers, but there >probably were some. That leaves the more common 24, 36, 48, and 60 >bit sizes. >Am I missing something? Yes, the lit'luns. 18bit: TX-0, pdp1 (first TECO), pdp4, pdp7 (first Unix), pdp9, pdp15 12bit: pdp5, LINC, pdp8, LINC-8, pdp12 In most of these (all the later ones) the addresses were equal wide as the data, instructions had shortend adresses allowing accessing a subset of memory directly, all above was accessed by indirect addressing. Of course orthogonal instruction sets had to be sacrificed. On the LINC only 3 instructions (ADD, STC (stor and clear acc), JMP (actually junp subroutine)) allowed direct 10bit adresses, rest direct 5bit, indirect 10bit. On the pdp8 only 6 instructions can access memory (AND, TAD (twos compl add), ISZ (incr skip if zero), DCA (deposit and clear acc), JSR, JMP) even can access memory (all direct 8bit, indirect 12bit). And yes, we know from previous postings that 1/4kword suffices. -- Neil.Franklin.remove.this@ccw.ch, http://www.ccw.ch/Neil.Franklin/ for Geek Code, Papernet, Voicenet, PGP public key see http: Mac, 95 and NT users are CLUEless (Command Line User Environment) If I go missing, its once again my newsfeed that has craped ###### From: "Joe Gorman" Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Wed, 28 Jan 1998 08:04:20 -0500 Organization: Medical University of South Carolina Lines: 4 Message-ID: <6ansaq$mf6@nellie.musc.edu> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aiadv$t2g$1@shadow.skypoint.net> NNTP-Posting-Host: 128.23.100.202 X-Newsreader: Microsoft Outlook Express 4.72.2106.4 X-MimeOLE: Produced By Microsoft MimeOLE V4.72.2106.4 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!newsxfer.visi.net!cpk-news-hub1.bbnplanet.com!atl-news-feed1.bbnplanet.com!news.bbnplanet.com!nellie.musc.edu!newsmaster Yes, the military had some 30 bit computers along with 18,24 and 32 bit machines, all of which I got to "play " with ###### From: "Joe Gorman" Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Wed, 28 Jan 1998 08:04:20 -0500 Organization: Medical University of South Carolina Lines: 4 Message-ID: <6ansaq$mf6@nellie.musc.edu> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aiadv$t2g$1@shadow.skypoint.net> NNTP-Posting-Host: 128.23.100.202 X-Newsreader: Microsoft Outlook Express 4.72.2106.4 X-MimeOLE: Produced By Microsoft MimeOLE V4.72.2106.4 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!newsxfer.visi.net!cpk-news-hub1.bbnplanet.com!atl-news-feed1.bbnplanet.com!news.bbnplanet.com!nellie.musc.edu!newsmaster Yes, the military had some 30 bit computers along with 18,24 and 32 bit machines, all of which I got to "play " with ###### Date: 28 Jan 98 17:41:02 -0800 From: "Charlie Gibbs" Subject: Re: word sizes (was: Re: Why IBM chose Intel) References: <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aiadv$t2g$1@shadow.skypoint.net> Message-ID: <3111.332T145T10613114@sky.bus.com> Newsgroups: alt.folklore.computers Lines: 66 X-Newsreader: THOR 2.5a (Amiga;TCP/IP) NNTP-Posting-Host: news.skybus.com Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news-out.internetmci.com!newsfeed.internetmci.com!206.229.87.25!news-peer.sprintlink.net!news-backup-east.sprintlink.net!news-in-east.sprintlink.net!news.sprintlink.net!Sprint!204.244.4.2!news.westel.com!news.skybus.com!204.244.247.128 In article tsw@cagent.com (Tom Watson) writes: >I once worked on a "graphics" computer, called an Adage. It used >30 bit words, and was programin Fortran of all things. Most of >the integer stuff was in "fractional" units, which meant that the >arithmetic was mostly 2's complement, but the decimal point was >on the left. In addition/subtraction this didn't mean much, but >on multiplication, you kept the MOST significant bits (15), not >the least significant ones. One of these beasts was at UBC when I was there (around 1970). I didn't officially have access to it, but a couple of us found it idle one day and came across enough manuals to enter a little program through the front panel that would shift pretty patterns across the lights. >One of the people using the machine worked up a "Lunar Lander" >goodie. It was quite interesting. You either landed the module, >or made a creater. If you landed sucessfully, you were greeted >with an American flag, and more if you had any "fuel" left. >Most of the time I made creaters!! Ours had an incarnation of the original Spacewar, complete with gravity and hyperspace. If you went off one side of the screen you'd wrap around to the other side, which made for some interesting effects with torpedoes. One time I got my ship going so fast that successive images were drawn a couple of inches apart on the screen. With the wraparound it made an interesting strobe effect. >The display was a stroke based CRT, not raster. It had "really big" >transistors to drive the yoke. The character generator was in software >(the machine didn't have a hardware generator). One of the things we >did was change the bootstrap from a chunk of paper tape (read in with >an ASR33) to a bunch of ROM's I got. Worked much better. If it was drawing a complex image, it would start out pretty fast and smooth, but would become progressively slower and more flickery. >The disks were also "interesting". It was a "dual" drive (made by >ICL??) that had two platters on one spindle. The removable disks >attached on the left and right side of the drive. You needed to put >them on differently depending upon which side you mounted the disk. Ours didn't have a disk drive; it had one or two Kennedy tape drives that I heard people cursing at a lot. >Pretty good text editor, mostly mode-less. The machine had all sorts >of "graphic" things. Box of pots, foot pedals, box of push buttons. >Didn't have a mouse, or trackball as they weren't really "invented" >(documented/wide use) when the machine was built (early 70's). Ours was interfaced to the 360/67 downstairs, and you'd download programs and data from your MTS account on the mainframe. One day someone discovered that after you downloaded Spacewar you could sign off the mainframe and play all night for free. The computer center staff eventually found out, and modified things so that you couldn't play offline anymore. :-( >Truly an interesting machine. Yup. -- cgibbs@sky.bus.com (Charlie Gibbs) Remove the first period after the "at" sign to reply. ###### From: gnagy@bigdog.fred.net Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: 4 Feb 1998 05:22:50 -0500 Organization: FredNet - Frederick, Md. Lines: 15 Message-ID: <6b9fhq$9i2@bigdog.fred.net> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> NNTP-Posting-Host: bigdog.fred.net X-Trace: news.fred.net 886588391 3306 204.215.84.2 (4 Feb 1998 10:33:11 GMT) X-Complaints-To: abuse@fred.net Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!dca1-hub1.news.digex.net!digex!news.fred.net!not-for-mail In article <34fb169d.464331033@206.210.64.12>, Joseph M. Newcomer wrote: >Why did so many machines, e.g., the 709, 7090, PDP-6, PDP-10, and >DEC-20 all have 36-bit words... The rumor travelling eastbound on Penn Ave is that this was foresight in planning so that one day, the Clinton white house would be able to count all the members of the vast right wing conspiracy. -- -- greg Eagles may soar, but weasels don't get sucked through jet engines ###### From: bo774@FreeNet.Carleton.CA (Kelly Bert Manning) Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: 4 Feb 1998 06:27:59 GMT Organization: The National Capital FreeNet Lines: 17 Message-ID: <6b91pf$8t3@freenet-news.carleton.ca> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aiadv$t2g$1@shadow.skypoint.net> <6aunu7$k81$1@yeppa.connect.com.au> Reply-To: bo774@FreeNet.Carleton.CA (Kelly Bert Manning) NNTP-Posting-Host: freenet3.carleton.ca X-Given-Sender: bo774@freenet3.carleton.ca (Kelly Bert Manning) Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!newspeer.monmouth.com!sunqbc.risq.qc.ca!news-peer.gip.net!news.gsl.net!gip.net!europa.clark.net!209.130.129.134!node2.frontiernet.net!node17.frontiernet.net!xcski.com!freenet-news.carleton.ca!FreeNet.Carleton.CA!bo774 The Thought Assassin (assassin@sleepless.south.networx.net.au) writes: > Sean Case wrote: >> In article <6aiadv$t2g$1@shadow.skypoint.net>, >> >I can't think offhand of any 30, 42, or 54 bit computers, but there >> >probably were some. >> The US military had some 30 bit machines, at least. (Probably still do.) > I suupose Deep Thought should have been 42 bit? > And if so, surely the Earth is likewise 54 bit? > Honeywell mainframes had 36 bit words, with byte sizes ranging from 4 to 9 bits. This had a truely bizarre consequence in their COBOL-74 implementation of group move. Their packed decimal was implemented as character arithmetic on 4 bit bytes, so if you did a group move of a group consisting of only comp-3 to a character or mixed group it would do a 4 bit to 9 bit conversion and not bother to tell you. ###### From: "George R. Gonzalez" Newsgroups: alt.folklore.computers Subject: Re: word sizes (was: Re: Why IBM chose Intel) Date: Wed, 4 Feb 1998 09:39:00 -0600 Organization: University of Minnesota Lines: 46 Message-ID: <6baacc$r9i@epx.cis.umn.edu> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aiadv$t2g$1@shadow.skypoint.net> <6aunu7$k81$1@yeppa.connect.com.au> <6b91pf$8t3@freenet-news.carleton.ca> NNTP-Posting-Host: b134-164.dhcp.umn.edu Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Newsreader: Microsoft Outlook Express 4.71.1712.3 X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news.he.net!Supernews60!supernews.com!news.eecs.umich.edu!newshub.tc.umn.edu!newsstand.tc.umn.edu!usenet Kelly Bert Manning wrote in message <6b91pf$8t3@freenet-news.carleton.ca>... > >The Thought Assassin (assassin@sleepless.south.networx.net.au) writes: >> Sean Case wrote: >>> In article <6aiadv$t2g$1@shadow.skypoint.net>, >>> >I can't think offhand of any 30, 42, or 54 bit computers, but there >>> >probably were some. >>> The US military had some 30 bit machines, at least. (Probably still do.) >> I suupose Deep Thought should have been 42 bit? >> And if so, surely the Earth is likewise 54 bit? >> >Honeywell mainframes had 36 bit words, with byte sizes ranging from >4 to 9 bits. This had a truely bizarre consequence in their COBOL-74 >implementation of group move. Their packed decimal was implemented as >character arithmetic on 4 bit bytes, so if you did a group move of >a group consisting of only comp-3 to a character or mixed group it >would do a 4 bit to 9 bit conversion and not bother to tell you. Could be worse. The CDC6600 series had 6-bit characters. But COBOL required that you could overpunch a minus sign over the first digit of a numeric field. But all 64 characters were already asigned in the character set-- there were not 10 spare characters to represent 0-9 with overpunch. Sooo the card-reader code converter would take numeric overpunches and map them to other printable characters (I forget which). Soooo they carefully did not document that a gibberish character in a numeric field would turn a number negative, with no warning! Much later they had even worse troubles, as the Govt required every computer to have at least 64 printable characters, and CDC had reserved the 00 character for end-of-line, leaving only 63 printable characters. They had to do the impossible and squeeze 65 characters into 6 bits. The software kludges involved almost worked but it was disheartening to have a $6,000,000 computer with a non-deterministic and unreliable character set. Regards, George ###### From: Eric Chomko Newsgroups: alt.folklore.computers Subject: Re: word sizes Date: 4 Feb 1998 20:28:01 GMT Organization: IDT Internet Services Lines: 13 Message-ID: <6baj0h$350@nnrp1.farm.idt.net> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aiadv$t2g$1@shadow.skypoint.net> <6as6ab$ma4@news1.panix.com> NNTP-Posting-Host: u1.farm.idt.net X-Newsreader: TIN [UNIX 1.3 unoff BETA release 961025] Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!newsfeed.usit.net!news.vt.edu!solaris.cc.vt.edu!newsrelay.netins.net!nntp.kreonet.re.kr!news.maxwell.syr.edu!newsxfer.visi.net!news.idt.net!nntp.farm.idt.net!u1.farm.idt.net!not-for-mail : >Joseph M. Newcomer wrote in message <34fb169d.464331033@206.210.64.12>... : >>OK, here's a question that will demonstrate if you are a newbie or a : >>genuine OF: : >> : >>Why did so many machines, e.g., the 709, 7090, PDP-6, PDP-10, and : >>DEC-20 all have 36-bit words... : >> joe : >>Joseph M. Newcomer Octal. Eric ###### From: Robert Billing Newsgroups: alt.folklore.computers Subject: Re: word sizes Date: Thu, 05 Feb 98 22:45:28 GMT Message-ID: <886718728snz@tnglwood.demon.co.uk> References: <885597899.5107.4.nnrp-06.9e98ee68@news.demon.co.uk> <6abjmq$6j8@mnementh.trin.cam.ac.uk> <34fb169d.464331033@206.210.64.12> <6aiadv$t2g$1@shadow.skypoint.net> <6as6ab$ma4@news1.panix.com> <6baj0h$350@nnrp1.farm.idt.net> Reply-To: unclebob@tnglwood.demon.co.uk X-Mail2News-User: unclebob@tnglwood.demon.co.uk X-Mail2News-Path: post-20.mail.demon.net!post.mail.demon.net!tnglwood.demon.co.uk X-Trace: mail2news.demon.co.uk 886722126 20833 unclebob tnglwood.demon.co.uk X-Complaints-To: abuse@demon.net X-Newsreader: Demon Internet Simple News v1.29 Lines: 13 Path: ccw.ch!aetna.dolphins.ch!news.planetc.com!news.igateway.net!news.pagesat.net!howland.erols.net!news-peer.gip.net!news-lond.gip.net!news.gsl.net!gip.net!dispose.news.demon.net!demon!news.demon.co.uk!demon!mail2news.demon.co.uk!tnglwood.demon.co.uk!unclebob In article <6baj0h$350@nnrp1.farm.idt.net> chomko@IDT.NET "Eric Chomko" writes: > Octal. Nice one. Your fish are in the post. -- I am Robert Billing, Christian, inventor, traveller, cook and animal lover, I live near 0:46W 51:22N. http://www.tnglwood.demon.co.uk/ "Bother," said Pooh, "Eeyore, ready two photon torpedoes and lock phasers on the Heffalump, Piglet, meet me in transporter room three"