Message-ID: <3997E0C4.BE19918@jetnet.ab.ca> Date: Mon, 14 Aug 2000 12:06:28 +0000 From: Ben Franchuk X-Mailer: Mozilla 4.73 [en] (X11; I; Linux 2.2.15 i586) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <2000Aug14.184433.8359@lorelei.approve.se> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit NNTP-Posting-Host: 207.153.6.59 X-Trace: 14 Aug 2000 12:17:22 -0700, 207.153.6.59 Organization: OA Internet Lines: 25 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.direct.ca!look.ca!west2.newsfeed.sprint-canada.net!news.oanet.com!207.153.6.59 Xref: chonsp.franklin.ch alt.folklore.computers:61871 alt.sys.pdp8:12 alt.sys.pdp10:52 alt.sys.pdp11:8 Goran Larsson wrote: > > In article , > Christian Bau wrote: > > > There is no real technical advantage of 8/16/32 bit. > > What about opcodes for things like bit test/set with the bit > number encoded as a bit field in the opcode? What should a > 12-bit processor do with these instructions when the bit field > (four bits) has a value between 12 and 15? Wasted space or used > as a completely different opcode? What did existing non (2^n)-bit > computers do in these situations? That could never happen - All 12 bit computers are coded in OCTAL. Really , 12 bit computers never had the OPCODE ROOM for bit test/set or multi-shifts. Most likely that would be a undefined opcode. Ben. -- "We do not inherit our time on this planet from our parents... We borrow it from our children." "Octal Computers:Where a step backward is two steps forward!" http://www.jetnet.ab.ca/users/bfranchuk/index.html ###### Message-ID: <3997E6E2.F39BA3EF@jetnet.ab.ca> Date: Mon, 14 Aug 2000 12:32:34 +0000 From: Ben Franchuk X-Mailer: Mozilla 4.73 [en] (X11; I; Linux 2.2.15 i586) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <2000Aug14.184433.8359@lorelei.approve.se> Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit NNTP-Posting-Host: 207.153.6.59 X-Trace: 14 Aug 2000 12:43:18 -0700, 207.153.6.59 Organization: OA Internet Lines: 29 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news-fra1.dfn.de!news0.de.colt.net!colt.net!news.maxwell.syr.edu!newsfeed.direct.ca!look.ca!west2.newsfeed.sprint-canada.net!news.oanet.com!207.153.6.59 Xref: chonsp.franklin.ch alt.folklore.computers:61876 alt.sys.pdp8:13 alt.sys.pdp10:53 alt.sys.pdp11:9 Goran Larsson wrote: > > In article , > Christian Bau wrote: > > > There is no real technical advantage of 8/16/32 bit. > > What about opcodes for things like bit test/set with the bit > number encoded as a bit field in the opcode? What should a > 12-bit processor do with these instructions when the bit field > (four bits) has a value between 12 and 15? Wasted space or used > as a completely different opcode? What did existing non (2^n)-bit > computers do in these situations? Checking the Net for 12 bit computers,I found the PDP-12 and the LINC computers do multi-shift of #n positions, but the documentation says nothing about values larger than 11. I assume it would just keep shifting. I don't know about CDC-160 computer. Ben. > -- > Göran Larsson Senior Systems Analyst hoh AT approve DOT se -- "We do not inherit our time on this planet from our parents... We borrow it from our children." "Octal Computers:Where a step backward is two steps forward!" http://www.jetnet.ab.ca/users/bfranchuk/index.html ###### From: "John Galt" Newsgroups: alt.folklore.computers References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> Subject: Re: Q: Why not (2^n)-bit? Lines: 11 X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2615.200 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2615.200 Message-ID: Date: Mon, 14 Aug 2000 14:41:29 GMT NNTP-Posting-Host: 24.31.7.150 X-Complaints-To: abuse@mediaone.net X-Trace: typhoon.mn.mediaone.net 966264089 24.31.7.150 (Mon, 14 Aug 2000 09:41:29 CDT) NNTP-Posting-Date: Mon, 14 Aug 2000 09:41:29 CDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.cwix.com!cyclone.columbus.rr.com!news.rr.com!cyclone-midwest.rr.com!cyclone.rdc-detw.rr.com!news.mw.mediaone.net!typhoon.mn.mediaone.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61921 Ben Hutchings wrote in message news:u3dk8m0dn.fsf@roundpoint.com... > The original ASCII code was 6-bit, and the later version is only > 7-bit. There may have been *talk* of a 6-bit ASCII during the proceedings that arrived at a code. But there was never an approved standard. ###### Message-ID: <399815F8.2BCA625F@etnus.com> From: James Cownie Organization: Etnus, Inc. X-Mailer: Mozilla 4.6 [en] (X11; I; Linux 2.2.16 i686) X-Accept-Language: en-GB,fr MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <8n8c13$1rk$1@pegasus.csx.cam.ac.uk> <8n9432$l66$1@pegasus.csx.cam.ac.uk> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 20 Date: Mon, 14 Aug 2000 15:54:42 GMT NNTP-Posting-Host: 213.48.99.166 X-Complaints-To: abuse@blueyonder.co.uk X-Trace: news1.cableinet.net 966268482 213.48.99.166 (Mon, 14 Aug 2000 16:54:42 BST) NNTP-Posting-Date: Mon, 14 Aug 2000 16:54:42 BST Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeeds.belnet.be!news.belnet.be!newsfeed.icl.net!ldn-newsfeed.speedport.net!newsfeed.speedport.net!news-hub.cableinet.net!news1.cableinet.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61869 Surely part of the reason arises from the switch to byte-addressing. (Whatever size byte you choose !) Once you have byte addressing, then it seems much more natural to have a power of 2 number of bytes per word. There are _many_ C programmers who assume that char * p = ...; (((long)p)+1) == (long)(p+1) and this, of course, is unlikely to hold on a machine with 3 bytes per word. (At least if it is implemented on what seems the most natural way in which not all integers are valid byte addresses...)  -- Jim James Cownie Etnus, LLC. +44 117 9071438 http://www.etnus.com ###### From: Keith Thompson Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 14 Aug 2000 10:45:51 -0700 Organization: CTSnet Internet Services Lines: 33 Sender: kst@king.cts.com Message-ID: References: <8n8c13$1rk$1@pegasus.csx.cam.ac.uk> <8n9432$l66$1@pegasus.csx.cam.ac.uk> <399815F8.2BCA625F@etnus.com> X-Trace: thoth.cts.com 966275156 5837 205.163.0.22 (14 Aug 2000 17:45:56 GMT) X-Complaints-To: abuse@cts.com X-Newsreader: Gnus v5.5/Emacs 20.3 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-was.dfn.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!nntp.flash.net!mercury.cts.com!thoth.cts.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61889 James Cownie writes: > Surely part of the reason arises from the switch to byte-addressing. > (Whatever size byte you choose !) > > Once you have byte addressing, then it seems much more natural > to have a power of 2 number of bytes per word. > > There are _many_ C programmers who assume that > char * p = ...; > > (((long)p)+1) == (long)(p+1) > > and this, of course, is unlikely to hold on a machine with > 3 bytes per word. (At least if it is implemented on what seems the most > natural way in which not all integers are valid byte addresses...) Really? Why would a C programmer make this assumption -- or more importantly, why would a C programmer write code that depends on this assumption? C provides a rich set of operations that work directly on pointers. There's simply no need, 99.9% of the time, to convert between pointers and integers. BTW, there are power-of-two machines where the assumption above doesn't hold (e.g., the Cray T90, where the offset portion of a byte pointer is stored in the high-order bits). I once ran into some code that depended on the assumption that pointer arithmetic could be done as integer arithmetic; after I fixed it, the resulting code was smaller and cleaner. -- Keith Thompson (The_Other_Keith) kst@cts.com San Diego Supercomputer Center <*> Welcome to the last year of the 20th century. ###### From: ab528@FreeNet.Carleton.CA (Heinz W. Wiggeshoff) Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: 14 Aug 2000 18:07:26 GMT Organization: The National Capital FreeNet Lines: 18 Message-ID: <8n9cgu$3jj$1@freenet9.carleton.ca> References: <8n8c13$1rk$1@pegasus.csx.cam.ac.uk> <8n9432$l66$1@pegasus.csx.cam.ac.uk> Reply-To: ab528@FreeNet.Carleton.CA (Heinz W. Wiggeshoff) NNTP-Posting-Host: freenet10 X-Trace: freenet9.carleton.ca 966276446 3699 134.117.136.30 (14 Aug 2000 18:07:26 GMT) X-Complaints-To: complaints@ncf.ca NNTP-Posting-Date: 14 Aug 2000 18:07:26 GMT X-Given-Sender: ab528@freenet10.carleton.ca (Heinz W. Wiggeshoff) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!news-stu1.dfn.de!news-koe1.dfn.de!news-was.dfn.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.stanford.edu!news.kjsl.com!xcski.com!freenet-news!FreeNet.Carleton.CA!ab528 Xref: chonsp.franklin.ch alt.folklore.computers:61900 Nick Maclaren (nmm1@cus.cam.ac.uk) writes: > > Not write system-dependencies into the code unless they were > necessary for its function or important for some other reason > and, even then, to localise them in a few, clean functions. > > With assumptions like this, you will often find that they are > completely unnecessary in 90%+ of the programs that make them, > in the sense that code that does not make the assumptions is > as concise, as clear and as efficient. Hmmm, this sounds familiar. Oh, yes - D.E.Knuth's The Art of Computer Programming. His description of the MIX mythical computer states that beyond certain minimal limits, word size and base are irrelevant, IIRC. ###### Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 From: hoh@invalid.invalid (Goran Larsson) Subject: Re: Q: Why not (2^n)-bit? Message-ID: <2000Aug14.184433.8359@lorelei.approve.se> Originator: hoh@invalid.invalid (Goran Larsson) Sender: hoh@lorelei.approve.se.NO_JUNK_EMAIL (Goran Larsson) Organization: [1] + 5934 done /bin/rm -rf ~/ & X-No-Archive: yes X-Newsreader: trn 4.0-test73 (May 24, 2000) References: <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> Lines: 14 NNTP-Posting-Host: 130.244.215.144 X-Complaints-To: news-abuse@swip.net X-Trace: nntpserver.swip.net 966279437 130.244.215.144 (Mon, 14 Aug 2000 20:57:17 MET DST) NNTP-Posting-Date: Mon, 14 Aug 2000 20:57:17 MET DST X-Sender: unknown@isdn215-1-144.swipnet.se Date: Mon, 14 Aug 2000 18:44:33 GMT 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!newsfeed.icl.net!newsfeed.icl.net!newspeer.clara.net!news.clara.net!newsfeed1.swip.net!swipnet!nntpserver.swip.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61904 alt.sys.pdp8:17 alt.sys.pdp10:57 alt.sys.pdp11:11 In article , Christian Bau wrote: > There is no real technical advantage of 8/16/32 bit. What about opcodes for things like bit test/set with the bit number encoded as a bit field in the opcode? What should a 12-bit processor do with these instructions when the bit field (four bits) has a value between 12 and 15? Wasted space or used as a completely different opcode? What did existing non (2^n)-bit computers do in these situations? -- Göran Larsson Senior Systems Analyst hoh AT approve DOT se ###### Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <2000Aug14.184433.8359@lorelei.approve.se> Organization: D Bit, Troy, NY From: wilson@dbit.com (John Wilson) NNTP-Posting-Host: dbit.dbit.com X-Original-NNTP-Posting-Host: dbit.dbit.com Message-ID: <39984e80$1_1@news.wizvax.net> Date: 14 Aug 2000 15:54:40 -0400 X-Trace: 14 Aug 2000 15:54:40 -0400, dbit.dbit.com Lines: 31 XPident: wilson X-Original-NNTP-Posting-Host: 199.181.141.3 XPident: news 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.seicom.net!news-in.ivm.net!newsfeed00.sul.t-online.de!newsfeed01.sul.t-online.de!t-online.de!news.netcologne.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newspeer.phoen-x.net!news.wizvax.net!dbit.com!wilson Xref: chonsp.franklin.ch alt.folklore.computers:61884 alt.sys.pdp8:14 alt.sys.pdp10:55 alt.sys.pdp11:10 In article <2000Aug14.184433.8359@lorelei.approve.se>, Goran Larsson wrote: >What about opcodes for things like bit test/set with the bit >number encoded as a bit field in the opcode? What should a >12-bit processor do with these instructions when the bit field >(four bits) has a value between 12 and 15? Well, the 12-bit CPUs I've used (PDP-8s) would respond by saying "guh?", since they have no hardware support for bit fields anyway. Lots of older CPUs didn't, and many of the ones that did still expected you to figure out for yourself which word your bit was in. So just divide by twelve (in software, most 8s had no divider or shift-by-N anyway) instead of sixteen and go from there -- that's how you'd want to address the bits anyway, if you need to be able to do math on the bit pointers. It's fun to pick on non-power-of-two word sizes as being ungainly and stupid, but honestly, most of the code you write doesn't deal with arbitrarily sized bit fields, or depend on how many shifts it takes to make something fall off the left end of the word. As long as the words/halfwords/whatever are wide enough to hold the numbers you want to use, you're all set. And having a word size that's way bigger than the address size can come in handy. Word size creap is really getting out of hand these days. IMHO, applications which "need" more than 4 GB of memory, probably "need" an infinite amount of memory, so going to 64 bits isn't going to be a permanent fix. But they'll hype everyone into buying new machines and software anyway, since the sheep all want to have the latest -- hmm, what was that other decade that had a lot of sixty-something-bit computers? :-) John Wilson D Bit ###### From: Eric Fischer Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: 14 Aug 2000 21:02:55 GMT Organization: EnterAct Corp. Lines: 19 Message-ID: <8n9mpv$1tct$1@news.enteract.com> References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> NNTP-Posting-Host: shell-1.enteract.com X-Trace: news.enteract.com 966286975 62877 207.229.143.40 (14 Aug 2000 21:02:55 GMT) X-Complaints-To: abuse@enteract.com NNTP-Posting-Date: 14 Aug 2000 21:02:55 GMT X-Newsreader: trn 4.0-test72 (19 April 1999) Originator: enf@enteract.com (Eric Fischer) 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!rz.uni-karlsruhe.de!schlund.de!newsfeed01.sul.t-online.de!newsfeed00.sul.t-online.de!t-online.de!diablo.theplanet.net!news.maxwell.syr.edu!newsfeed.enteract.com!news.enteract.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61899 John Galt wrote: > Ben Hutchings wrote in message > > > The original ASCII code was 6-bit, and the later version is only > > 7-bit. > > There may have been *talk* of a 6-bit ASCII during the proceedings that > arrived at a code. But there was never an approved standard. Not in the USA, at least. But while there was never a 6-bit code called ASCII, there *was* a compatible 6-bit code ECMA-1 from the European Computer Manufacturers' Association. I believe the first publication of ISO R-646 included both 6- and 7-bit versions. In the 7-bit codes, the most important controls are clustered in the bottom half of the first row of the table to facilitate conversions between the 6- and 7-bit versions of the code. eric ###### Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 From: hoh@invalid.invalid (Goran Larsson) Subject: Re: Q: Why not (2^n)-bit? Message-ID: <2000Aug14.212932.9879@lorelei.approve.se> Originator: hoh@invalid.invalid (Goran Larsson) Sender: hoh@lorelei.approve.se.NO_JUNK_EMAIL (Goran Larsson) Organization: [1] + 5934 done /bin/rm -rf ~/ & X-No-Archive: yes X-Newsreader: trn 4.0-test73 (May 24, 2000) References: <2000Aug14.184433.8359@lorelei.approve.se> <39984e80$1_1@news.wizvax.net> Lines: 26 NNTP-Posting-Host: 130.244.215.104 X-Complaints-To: news-abuse@swip.net X-Trace: nntpserver.swip.net 966289010 130.244.215.104 (Mon, 14 Aug 2000 23:36:50 MET DST) NNTP-Posting-Date: Mon, 14 Aug 2000 23:36:50 MET DST X-Sender: unknown@isdn215-1-104.swipnet.se Date: Mon, 14 Aug 2000 21:29:32 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!newsfeedZ.netscum.dQ!netscum.int!news.algonet.se!algonet!newsfeed1.swip.net!swipnet!nntpserver.swip.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61949 alt.sys.pdp8:21 alt.sys.pdp10:60 alt.sys.pdp11:14 In article <39984e80$1_1@news.wizvax.net>, John Wilson wrote: > Well, the 12-bit CPUs I've used (PDP-8s) would respond by saying "guh?", > since they have no hardware support for bit fields anyway. Lots of older CPUs > didn't, and many of the ones that did still expected you to figure out for > yourself which word your bit was in. I wasn't writing about bitfields as used by the programmer. I was writing about the bits inside the OPCODE. If you have an instruction on a 12-bit machine that can do something with a single bit in an register the opcode must use some bits to address the bit. As three bits are too few (enough for a 8-bit register) and four bits are too many (enough for a 16-bit register) they had to use four bits and waste some combinations or use the leftover combinations for something else. > It's fun to pick on non-power-of-two word sizes as being ungainly and stupid, > but honestly, most of the code you write doesn't deal with arbitrarily sized > bit fields, or depend on how many shifts it takes to make something fall off Forget the code at user level. What is done inside the hardware? Using a wordlength not equal to (2^n)-bit for a binary machine seems, to me, not to be very elegant. It saved memory and hardware many years ago, but today it is just ugly. -- Göran Larsson Senior Systems Analyst hoh AT approve DOT se ###### From: pw@panix.com (Paul Wallich) Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8 Subject: Re: Q: Why not (2^n)-bit? Date: Mon, 14 Aug 2000 18:17:45 -0400 Organization: PANIX Public Access Internet and UNIX, NYC Lines: 20 Message-ID: References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8n91sd$107g$1@ausnews.austin.ibm.com> <39985A75.485B@hda.hydro.com> NNTP-Posting-Host: pw.dialup.access.net X-Trace: news.panix.com 966291467 27408 166.84.250.178 (14 Aug 2000 22:17:47 GMT) X-Complaints-To: abuse@panix.com NNTP-Posting-Date: 14 Aug 2000 22:17:47 GMT X-Newsreader: MT-NewsWatcher 2.4.4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!howland.erols.net!panix!news.panix.com!pw Xref: chonsp.franklin.ch alt.folklore.computers:61962 alt.sys.pdp8:25 In article <39985A75.485B@hda.hydro.com>, Terje Mathisen wrote: >glass2@glass2.lexington.ibm.com wrote: >> But, 12 bits isn't enough to handle all of the possible character sets. >> Take a look at the Unicode specifications to see what's happening in the >> world of character set development (32 bit characters?). > >UTF8 defines encodings for up to 31-bit chars, so presumably the last 2G >of encodings aren't used. > >"What, only 2e9 different characters to choose from? That's less than >one per human being!" Which means, I guess, that we can't all be named Prince... More likely, I would expect that most of that space is effectively taken up with encodings that would blow up with framing errors etc. paul ###### From: "Stephen Fuld" Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 References: <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <2000Aug14.184433.8359@lorelei.approve.se> Subject: Re: Q: Why not (2^n)-bit? Lines: 47 X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2919.6600 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6600 Message-ID: <4j_l5.8371$4T.488747@bgtnsc07-news.ops.worldnet.att.net> Date: Mon, 14 Aug 2000 22:22:56 GMT NNTP-Posting-Host: 12.72.161.137 X-Complaints-To: abuse@worldnet.att.net X-Trace: bgtnsc07-news.ops.worldnet.att.net 966291776 12.72.161.137 (Mon, 14 Aug 2000 22:22:56 GMT) NNTP-Posting-Date: Mon, 14 Aug 2000 22:22:56 GMT Organization: AT&T Worldnet 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!europa.netcrusader.net!204.127.161.3!wn3feed!worldnet.att.net!wnmasters3!bgtnsc07-news.ops.worldnet.att.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61944 alt.sys.pdp8:19 alt.sys.pdp10:59 alt.sys.pdp11:13 "Goran Larsson" wrote in message news:2000Aug14.184433.8359@lorelei.approve.se... > In article , > Christian Bau wrote: > > > There is no real technical advantage of 8/16/32 bit. > > What about opcodes for things like bit test/set with the bit > number encoded as a bit field in the opcode? What should a > 12-bit processor do with these instructions when the bit field > (four bits) has a value between 12 and 15? Wasted space or used > as a completely different opcode? What did existing non (2^n)-bit > computers do in these situations? I certainly can't speak for all of them but the Univac/Sperry/Unisys 1100/2200 series are 36 bit word addressed machines, originally designed in the mid 1960s, with models still in production. It has no direct bit test instruction, but has an interesting method of accomplishing the same thing. There are Test Even Parity and Test Odd Parity instructions. these instructions bitwise AND the contents of a specified register with another operand (which could be another register, a memory location, or an 18 bit immediate filed) and skipped the next instruction if the result of the AND had even or odd respectivly parity. The commmon way of testing a for a bit in a word in a register is to use these instructions with the second operand consisting of all zero bits except for the one you want to test for. If you used Test Odd Parity, the instruction would skip if the bit was set. Unfortunately, the only way to set a bit was the bitwise OR with a word of all zeros except the bit you wanted set. Of course, you frequently had that constant around anyway from the test instruction :-). And if the bit was in the low order 18 bits, you could use the immediate field.. -- - Stephen Fuld > > -- > Göran Larsson Senior Systems Analyst hoh AT approve DOT se ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. X-Newsreader: Gnus v5.5/Emacs 20.3 Date: 14 Aug 2000 15:58:02 -0700 Message-ID: Lines: 25 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 14 Aug 2000 16:02:50 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:61953 alt.sys.pdp8:23 alt.sys.pdp10:62 alt.sys.pdp11:16 In article , dls2 wrote: > PDP-01 1960-vintage 18-bit > PDP-02 19??-vintage 24-bit > PDP-03 19??-vintage 36-bit > PDP-04 1962-vintage 18-bit > PDP-05 1963-vintage 12-bit > PDP-06 1964-vintage 36-bit > PDP-07 1965-vintage 18-bit > PDP-08 1965-vintage 12-bit > PDP-09 1966-vintage 18-bit > PDP-10 1967-vintage 36-bit > PDP-11 1970-vintage 16-bit > PDP-12 1969-vintage 12-bit > PDP-13 19??-vintage ??-bit > PDP-14 19??-vintage ??-bit > PDP-15 1970-vintage 18-bit > PDP-16 1972-vintage ??-bit wb@yorikke.arb-phys.uni-dortmund.de (Wilhelm B. Kloke) writes: > Perhaps you did notice that all bit widths are either 2^n or 6*2^n. 18-bit is neither 2^n nor 6*2^n. (For integer values of n.) Perhaps you mean "2^n or 6*n". ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. X-Newsreader: Gnus v5.5/Emacs 20.3 Date: 14 Aug 2000 15:59:11 -0700 Message-ID: Lines: 6 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 14 Aug 2000 16:03:59 -0700, ruckus.brouhaha.com 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!feeder.via.net!enews.sgi.com!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:61954 alt.sys.pdp8:24 alt.sys.pdp10:63 alt.sys.pdp11:17 Ben Hutchings writes: > The original ASCII code was 6-bit, No, ASCII has *always* been at least 7 bits per character. Before ASCII and EBCDIC, many of the character codes were 5 or 6 bit. ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <39a013e5.25745326@news.remarq.com> <39982258.FCC144BF@etnus.com> <39984CFD.F1B2CA64@srv.net> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 14 Aug 2000 16:11:17 -0700 Message-ID: Lines: 15 X-Newsreader: Gnus v5.5/Emacs 20.3 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 14 Aug 2000 16:16:06 -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.folklore.computers:61952 alt.sys.pdp8:22 alt.sys.pdp10:61 alt.sys.pdp11:15 dls2 wrote: > Why then, were VAXen chosen to be 32-bit, and not 36-bit. Kevin Handy writes: > I wonder if the available memory technology had something to do > with it. Most of the memory chips were developed with the idea > of interfacing to the 80x86, which is a 2^n based system. *NONE* of the memory chips that were available when the VAX was being designed were specific to the 80x86 in any way. Especially considering that the VAX-11/780 was introduced in 1977, and the 8086 in 1978. The commonly available RAM chips in 1977 were SRAMs in 1-bit and 4-bit wide configurations, and DRAMs in 1-bit wide. These all work equally well for 32-bit or 36-bit word lengths. ###### Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <2000Aug14.184433.8359@lorelei.approve.se> <39984e80$1_1@news.wizvax.net> <2000Aug14.212932.9879@lorelei.approve.se> Organization: D Bit, Troy, NY From: wilson@dbit.com (John Wilson) NNTP-Posting-Host: dbit.dbit.com X-Original-NNTP-Posting-Host: dbit.dbit.com Message-ID: <399895cd$1_1@news.wizvax.net> Date: 14 Aug 2000 20:58:53 -0400 X-Trace: 14 Aug 2000 20:58:53 -0400, dbit.dbit.com Lines: 55 XPident: wilson X-Original-NNTP-Posting-Host: 199.181.141.3 XPident: news 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!newsfeed.cwix.com!news.mv.net!newspeer.phoen-x.net!news.wizvax.net!dbit.com!wilson Xref: chonsp.franklin.ch alt.folklore.computers:61941 alt.sys.pdp8:18 alt.sys.pdp10:58 alt.sys.pdp11:12 In article <2000Aug14.212932.9879@lorelei.approve.se>, Goran Larsson wrote: >I wasn't writing about bitfields as used by the programmer. I was >writing about the bits inside the OPCODE. If you have an instruction >on a 12-bit machine that can do something with a single bit in an >register the opcode must use some bits to address the bit. As three >bits are too few (enough for a 8-bit register) and four bits are too many >(enough for a 16-bit register) they had to use four bits and waste >some combinations or use the leftover combinations for something else. Again you're thinking about modern machines. The PDP-8 (dunno about other 12-bit machines like the ND-812 or whatever else there was) didn't have any instructions which took single-bit operands, except for the ones which operated on the LINK bit (like a carry bit), and there's only one of those. There aren't enough bits to have any of the modern-day fanciness of encoding complicated operands into the opcode, and even if they were the 8 wouldn't have used them, since it was designed to be a cheap Model T type computer. The PDP-8 didn't even have an immediate mode! So encoding immediate bit values in a hypothetical "bit test" opcode wouldn't be fair. In any case it didn't have a multi-shifter (except as an option and in that case you'd load the step count register as a separate operation), it could rotate only by one or two bits (or byte swap in later models), so you had to shake the bit you wanted into the LINK with RARs and RTRs before you could do anything with it. >Forget the code at user level. What is done inside the hardware? >Using a wordlength not equal to (2^n)-bit for a binary machine seems, >to me, not to be very elegant. What I was saying was, sure you can think of cases where it's a pain, but that only accounts for maybe 5% of regular everyday programming. Most of the time, you're operating on *numbers* (or characters), not bits, so you wouldn't even notice if it was actually a BCD machine or 19-bit Fibonacci machine or whatever. You still get 2+2=4 no matter how you look at it. A much bigger problem with the PDP-8 (and probably other 12-bit machines too) is the fact that the tiny word size requires a very small instruction set (unless you mess around with prefix opcodes, but that's more complexity than is appropriate for a low-end machine). The thing has no subtract or compare instructions, or even "load", you have to do all that with adds! And there's no hardware support for stacks, so you can't do recursion w/o adding a bunch of extra code. And yet, if you stop fighting it and pick algorithms that fit the hardware instead of the other way around, even the lobotomized instruction set isn't as bad as it seems, really it's pretty natural once you get used to it. It has only 8 basic instructions, but they're the *right* 8 instructions! >It saved memory and hardware many years >ago, but today it is just ugly. As someone already said another way, "modern" CPUs have found ways to be ugly which far surpass word size. It's just not that big a deal. John Wilson D Bit ###### From: jmaynard@thebrain.conmicro.cx (Jay Maynard) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 15 Aug 2000 13:04:39 GMT Organization: Neosoft (using Airnews.net!) Lines: 10 Message-ID: X-Orig-Message-ID: References: <8n8c13$1rk$1@pegasus.csx.cam.ac.uk> <8n9432$l66$1@pegasus.csx.cam.ac.uk> <399815F8.2BCA625F@etnus.com> Reply-To: jmaynard@conmicro.cx Abuse-Reports-To: abuse at airmail.net to report improper postings NNTP-Proxy-Relay: library2.airnews.net NNTP-Posting-Time: Tue Aug 15 08:04:39 2000 NNTP-Posting-Host: !cTJ31k-XT'Ukqe (Encoded at Airnews!) User-Agent: slrn/0.9.5.4 (UNIX) 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!howland.erols.net!news-out.nntp.airnews.net.MISMATCH!cabal10.airnews.net!news.airnews.net!cabal14.airnews.net!news.airnews.net!cabal2.airnews.net!news-f.iadfw.net!jmaynard Xref: chonsp.franklin.ch alt.folklore.computers:61943 On 14 Aug 2000 10:45:51 -0700, Keith Thompson wrote: >Really? Why would a C programmer make this assumption -- or more >importantly, why would a C programmer write code that depends on this >assumption? C provides a rich set of operations that work directly >on pointers. There's simply no need, 99.9% of the time, to convert >between pointers and integers. I have absolutely no idea why C programmers constantly assume that pointers and integers are interchangeable, but there's an obscene amount of code out there that does exactly that. Don't believe me? Get an Alpha. ###### Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <8nbbvn$1fks$1@citadel.in.taronga.com> Organization: Massachvsetts Institvte of Technology From: jfc@mit.edu (John F Carr) Date: 15 Aug 2000 13:07:26 GMT Lines: 18 Message-ID: <3999408e$0$9424@senator-bedfellow.mit.edu> NNTP-Posting-Host: NERD-XING.MIT.EDU X-Trace: dreaderd 966344846 9424 18.184.0.47 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!newsswitch.lcs.mit.edu!bloom-beacon.mit.edu!senator-bedfellow.mit.edu!dreaderd!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61955 In article <8nbbvn$1fks$1@citadel.in.taronga.com>, Peter da Silva wrote: >That doesn't mean that any arbitrary selection of 2^20 colors is going to >be enough to cover those 2^20 values, since the response of the eye doesn't >have any direct mapping to electron gun intensity levels and phosphor >brightness. Lookup tables or functions are common when image quality matters. I worked on a medical printer that accepted 8 bit input. The image on the film after going through a configurable LUT was 10-12 bits at approximately equal logarithmic intensity intervals ("density"). The later model had a 12 bit to 12 bit LUT. High end video devices have similar features (e.g. gamma correction or LUT to make 8 bit data produce 2^8 distinct intensities). -- John Carr (jfc@mit.edu) ###### Newsgroups: alt.folklore.computers,comp.arch References: <399815F8.2BCA625F@etnus.com> X-No-Archive: yes Organization: Crystal Cave Subject: Re: Q: Why not (2^n)-bit? X-Newsreader: trn 4.0-test72 (19 April 1999) From: plugh@NO.SPAM.PLEASE (Caveman) Lines: 24 Message-ID: Date: Tue, 15 Aug 2000 16:23:15 GMT NNTP-Posting-Host: 216.61.138.75 X-Complaints-To: abuse@swbell.net X-Trace: nnrp2.sbc.net 966356595 216.61.138.75 (Tue, 15 Aug 2000 11:23:15 CDT) NNTP-Posting-Date: Tue, 15 Aug 2000 11:23:15 CDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!skynet.be!newspush.london1.eu.level3.net!level3eu!newsfeed.mathworks.com!cyclone.swbell.net!nnrp2.sbc.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61942 In article , Jay Maynard wrote: > >I have absolutely no idea why C programmers constantly assume that pointers >and integers are interchangeable, but there's an obscene amount of code out >there that does exactly that. Don't believe me? Get an Alpha. In my experience, the most common reason has always been that few C programmers seem to really understand what a pointer is. This often leads them to substitute integer arithmetic, in which they have a higher confidence level, though still little real understanding of how this is implemented at runtime. Programmers, in other words, show a certain fear and prejudice toward pointer ops because they seem mysterious and strange. It might be fruitful to develop a programmer psychologist who could guide the psychological aspects of programmer mindset, just so long as it's not a dancing paperclip. ;) -- "You are in a maze of twisty little passages, all different." ###### Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 From: hoh@invalid.invalid (Goran Larsson) Subject: Re: Q: Why not (2^n)-bit? Message-ID: <2000Aug15.194425.18813@lorelei.approve.se> Originator: hoh@invalid.invalid (Goran Larsson) Sender: hoh@lorelei.approve.se.NO_JUNK_EMAIL (Goran Larsson) Organization: [1] + 5934 done /bin/rm -rf ~/ & X-No-Archive: yes X-Newsreader: trn 4.0-test73 (May 24, 2000) References: <39994EC4.107425B4@netinsight.se> <39997166.680536C6@honeywell.com> Lines: 11 NNTP-Posting-Host: 130.244.215.96 X-Complaints-To: news-abuse@swip.net X-Trace: nntpserver.swip.net 966368808 130.244.215.96 (Tue, 15 Aug 2000 21:46:48 MET DST) NNTP-Posting-Date: Tue, 15 Aug 2000 21:46:48 MET DST X-Sender: unknown@isdn215-1-96.swipnet.se Date: Tue, 15 Aug 2000 19:44:25 GMT 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-ulm.de!rz.uni-karlsruhe.de!schlund.de!newsfeed01.sul.t-online.de!newsfeed00.sul.t-online.de!t-online.de!diablo.theplanet.net!newsfeed1.swip.net!swipnet!nntpserver.swip.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62008 alt.sys.pdp8:30 alt.sys.pdp10:71 alt.sys.pdp11:23 In article <39997166.680536C6@honeywell.com>, Dan Thompson wrote: > I did not hear about the AMD2900 4-bit slice ICs until a few years > later. I would have to look up stuff at home to pin-point the time for > this. DEC used these critters in the VAX 11/730. -- Göran Larsson Senior Systems Analyst hoh AT approve DOT se ###### Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 From: hoh@invalid.invalid (Goran Larsson) Subject: Re: Q: Why not (2^n)-bit? Message-ID: <2000Aug15.194754.18934@lorelei.approve.se> Originator: hoh@invalid.invalid (Goran Larsson) Sender: hoh@lorelei.approve.se.NO_JUNK_EMAIL (Goran Larsson) Organization: [1] + 5934 done /bin/rm -rf ~/ & X-No-Archive: yes X-Newsreader: trn 4.0-test73 (May 24, 2000) References: <39997489$0$9439@senator-bedfellow.mit.edu> Lines: 13 NNTP-Posting-Host: 130.244.215.16 X-Complaints-To: news-abuse@swip.net X-Trace: nntpserver.swip.net 966373008 130.244.215.16 (Tue, 15 Aug 2000 22:56:48 MET DST) NNTP-Posting-Date: Tue, 15 Aug 2000 22:56:48 MET DST X-Sender: unknown@isdn215-1-16.swipnet.se Date: Tue, 15 Aug 2000 19:47:54 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!unlisys!news.snafu.de!newsfeedZ.netscum.dQ!netscum.int!news.algonet.se!algonet!newsfeed1.swip.net!swipnet!nntpserver.swip.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62009 alt.sys.pdp8:31 alt.sys.pdp10:72 alt.sys.pdp11:24 In article <39997489$0$9439@senator-bedfellow.mit.edu>, John F Carr wrote: > I thought the VAX 11/750 was the bit slice CPU. This is what I remember from a previous life in VAX-land. VAX 11/780 TTL VAX 11/750 gate arrays VAX 11/730 AMD 2900 -- Göran Larsson Senior Systems Analyst hoh AT approve DOT se ###### Message-ID: <39999FD2.A424BCBB@jetnet.ab.ca> Date: Tue, 15 Aug 2000 19:53:54 +0000 From: Ben Franchuk X-Mailer: Mozilla 4.73 [en] (X11; I; Linux 2.2.15 i586) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <3999892D.6B8CC2E3@trailing-edge.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit NNTP-Posting-Host: 207.153.6.46 X-Trace: 15 Aug 2000 16:46:40 -0700, 207.153.6.46 Organization: OA Internet Lines: 24 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!logbridge.uoregon.edu!newsfeed.direct.ca!look.ca!west2.newsfeed.sprint-canada.net!news.oanet.com!207.153.6.46 Xref: chonsp.franklin.ch alt.folklore.computers:61977 alt.sys.pdp8:27 alt.sys.pdp10:65 alt.sys.pdp11:19 Tim Shoppa wrote: > > Was there ever an architecture where opcode bits mapped *directly* > into 74181 control inputs? A lot of the 74181 control inputs > produced rather silly results, so this would seem to me to be a > waste of opcode bits at a time when memory was very expensive, but > if someone wants to point out a successful or unsuccesful > architecture that pulled this trick, I'd appreciate learning about it! > > Tim. The paper - "A minimal TTL Processor for architecture exploration" by Bradford J Rodriguez does that. I believe BYTE had a TTL processor that used a 74181. Check out http://www.ulib.org/webRoot/Books/Saving_Bell_Books/ The online book Computer Engineering. It has some good information on DEC's computer lines to about 1975. Ben. PS. My home-brew computer uses 74LS382's and the opcodes reflect the layout of that chip. --- "We do not inherit our time on this planet from our parents... We borrow it from our children." "Octal Computers:Where a step backward is two steps forward!" http://www.jetnet.ab.ca/users/bfranchuk/index.html ###### X-Newsreader: xrn 9.00 From: michael.wojcik@merant.com (Michael Wojcik) Reply-To: michael.wojcik@merant.com Subject: Re: Q: Why not (2^n)-bit? Newsgroups: alt.folklore.computers Followup-To: alt.folklore.computers References: <8n8c13$1rk$1@pegasus.csx.cam.ac.uk> <8n9432$l66$1@pegasus.csx.cam.ac.uk> <399815F8.2BCA625F@etnus.com> Organization: MERANT Inc. Originator: mww@lorelei.michaelwojcik.org Lines: 24 Message-ID: Date: Tue, 15 Aug 2000 20:17:16 GMT NNTP-Posting-Host: 158.252.173.87 X-Complaints-To: abuse@earthlink.net X-Trace: newsread2.prod.itd.earthlink.net 966370636 158.252.173.87 (Tue, 15 Aug 2000 13:17:16 PDT) NNTP-Posting-Date: Tue, 15 Aug 2000 13:17:16 PDT 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!logbridge.uoregon.edu!news.maxwell.syr.edu!newsfeed1.earthlink.net!newsmaster1.prod.itd.earthlink.net!newsread2.prod.itd.earthlink.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61990 [Dropping comp.arch.] In article , jmaynard@thebrain.conmicro.cx (Jay Maynard) writes: > I have absolutely no idea why C programmers constantly assume that pointers > and integers are interchangeable, but there's an obscene amount of code out > there that does exactly that. Don't believe me? Get an Alpha. Or an AS/400, where the firmware invalidates a pointer that's manipulated illegally (eg. by writing into it with memcpy) and where there is no C integer type that's large enough to contain the data in a C pointer type. -- Michael Wojcik michael.wojcik@merant.com AAI Development, MERANT (block capitals are a company mandate) Department of English, Miami University Unlikely predition o' the day: Eventually, every programmer will have to write a Java or distributed object program. -- Orfali and Harkey, _Client / Server Programming with Java and CORBA_ ###### From: torek@elf.bsdi.com (Chris Torek) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Organization: none of the above Lines: 69 Message-ID: <8nca8j$3vt$1@elf.bsdi.com> References: <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8n9n30$1tsg$1@news.enteract.com> Date: Tue, 15 Aug 2000 20:47:32 GMT NNTP-Posting-Host: 63.193.109.34 X-Complaints-To: abuse@flash.net X-Trace: news.flash.net 966372452 63.193.109.34 (Tue, 15 Aug 2000 15:47:32 CDT) NNTP-Posting-Date: Tue, 15 Aug 2000 15:47:32 CDT 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!nntp.flash.net!news.flash.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61974 (I deleted a bunch of now-obviously-irrelevant groups.) >dls2 wrote: >> "IBM PC Extended ASCII Display Characters >> Strictly speaking, the ASCII character set only includes values >> up to 127 decimal (7F hex). However, when the IBM PC was >> developed, the video card contained one byte for each character >> in the 80x25 character display. ... >> >> If this is true, then why did the video card contain one (8-bit) byte >> for each character? In article <8n9n30$1tsg$1@news.enteract.com>, Eric Fischer wrote: >Because all the other memory was 8 bits wide (actually, 8 bits >plus a parity bit), so it would have been more trouble than it >was worth to have a range of 7-bit wide memory just for the >display. > >But note that not all computer designers made the same decision: >The original TRS-80 Model I, for instance, had only 7-bit-wide >video memory, and used half of the range for the uppercase-only >subset of ASCII and the other half for block graphics. Quite true. Note also that the TRS-80 Model I was designed before the IBM PC. The video circuit in the TRS-80 consisted of seven 21L02 (static RAM) chips plus some extra logic using "leftover" gates on existing TTL chips on the main board. The extra logic computed data-bit 6 "on the fly" based on other bits, and fed the result to the character generator. Adding an eighth RAM chip would have increased the parts count, and hence the cost. In other words, different design constraints and cost functions led to different results. The character generator inside the (or at least "my") TRS-80 Model I did have a full set of lowercase characters. I used the "piggyback a new 21L02 on top of an existing one, with the data pin bent out to solder a wire to it" method to add lowercase to mine. I then discovered that the Microsoft ROM used some amazingly ugly code to write byte values to the video RAM. The code essentially did this: putchar(BYTE c) { BYTE *video_ram_pointer; ... c -= 64; if (c >= 0) { /* must have been alphabetic */ c -= 32; if (c < 0) /* must have been lowercase */ c += 32; *video_ram_pointer = c; goto update_cursor_position; } ... take care of other characters ... } The effect was that all alphabetic characters were stored to memory as bytes in the range [0..31]. Once I actually had lowercase, they would display as garbage. They relied on the fact that values in that range would "read" as values in the range [64..95] instead. (Fortunately, I had decided to add an extra few wires and a switch in the back of the case, so that I could turn lowercase off. That let me see what I was doing long enough to write a lowercase video driver.) -- In-Real-Life: Chris Torek, Berkeley Software Design Inc El Cerrito, CA, USA Domain: torek@bsdi.com +1 510 234 3167 http://claw.bsdi.com/torek/ (not always up) I report spam to abuse@. ###### From: Eric Fischer Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 15 Aug 2000 21:21:41 GMT Organization: EnterAct Corp. Lines: 27 Message-ID: <8ncc95$1bn2$1@news.enteract.com> References: <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8n9n30$1tsg$1@news.enteract.com> <8nca8j$3vt$1@elf.bsdi.com> NNTP-Posting-Host: shell-1.enteract.com X-Trace: news.enteract.com 966374501 44770 207.229.143.40 (15 Aug 2000 21:21:41 GMT) X-Complaints-To: abuse@enteract.com NNTP-Posting-Date: 15 Aug 2000 21:21:41 GMT X-Newsreader: trn 4.0-test72 (19 April 1999) Originator: enf@enteract.com (Eric Fischer) 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!newsfeed.enteract.com!news.enteract.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61995 Chris Torek wrote: > The video circuit in the TRS-80 consisted of seven > 21L02 (static RAM) chips plus some extra logic using "leftover" > gates on existing TTL chips on the main board. The extra logic > computed data-bit 6 "on the fly" based on other bits, and fed the > result to the character generator. > > Adding an eighth RAM chip would have increased the parts count, > and hence the cost. In other words, different design constraints > and cost functions led to different results. > > The character generator inside the (or at least "my") TRS-80 Model I > did have a full set of lowercase characters. Didn't they still manage to squeeze a few extra cents out of the character generator, by choosing one with such badly designed lowercase letters that no one who actually wanted to use lowercase would choose it? (I think one of the letters was upside down or something like that.) I think if you bought the upgrade from Radio Shack, they gave you a new character generator ROM too, but if you did it yourself you had to put up with the broken one. Thanks for posting the details of how the ROM expected the hardware to work. eric ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. X-Newsreader: Gnus v5.5/Emacs 20.3 Date: 15 Aug 2000 14:24:22 -0700 Message-ID: Lines: 8 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 15 Aug 2000 14:29:21 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!psinet-eu-nl!psiuk-p4!uknet!nntp.flash.net!enews.sgi.com!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62010 alt.sys.pdp8:32 alt.sys.pdp10:73 alt.sys.pdp11:25 nmm1@cus.cam.ac.uk (Nick Maclaren) writes: > Hmm. What species does he belong to? Mere humans cannot distinguish > more than about 2^20 colour/intensity combinations .... In a single scene. But the human eye (and most photographic film) have a *very* large dynamic range, many orders of magnitude. This is why MPEG with 8-bit quantization per color element is NOT an acceptable substitute for motion picture film. ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. X-Newsreader: Gnus v5.5/Emacs 20.3 Date: 15 Aug 2000 14:30:07 -0700 Message-ID: Lines: 13 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 15 Aug 2000 14:35:05 -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.mathworks.com!enews.sgi.com!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62013 alt.sys.pdp8:33 alt.sys.pdp10:74 alt.sys.pdp11:26 Dan.Pop@cern.ch (Dan Pop) writes: > Later PDP-11 models and the low cost VAX-11/730 were built using AMD 2900 > parts. The only place I remember seeing a 2900 in a PDP-11 was the floating point board for the 11/44. In most [*] of the TTL implementations, the CPU data path used 74181 or 74S181 ALU chips. Eric [*] I'm not sure about the 11/20. ###### Sender: Max Jester From: Max Jester Subject: Re: Q: Why not (2^n)-bit? Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 References: User-Agent: tin/1.4.3-20000502 ("Marian") (UNIX) (SunOS/5.7 (sun4u)) Lines: 13 Message-ID: Date: Tue, 15 Aug 2000 21:57:26 GMT NNTP-Posting-Host: 147.25.48.31 X-Complaints-To: news@ext.ray.com X-Trace: bos-service2.ext.raytheon.com 966376646 147.25.48.31 (Tue, 15 Aug 2000 17:57:26 EDT) NNTP-Posting-Date: Tue, 15 Aug 2000 17:57:26 EDT Organization: Raytheon Company 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!cpk-news-hub1.bbnplanet.com!cambridge1-snf1.gtei.net!news.gtei.net!bos-service1.ext.raytheon.com!bos-service2.ext.raytheon.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61970 alt.sys.pdp8:26 alt.sys.pdp10:64 alt.sys.pdp11:18 In alt.sys.pdp10 Eric Smith wrote: > The only place I remember seeing a 2900 in a PDP-11 was the floating point > board for the 11/44. In most [*] of the TTL implementations, the CPU > data path used 74181 or 74S181 ALU chips. Um, are you sure about that? I recall when we popped the FP board into an 11/44 and ISTR that the board consisted of 16 74181s plus TTL glue. HTH Max -- Disclaimer: I speak for no one but myself. ###### Message-ID: <3999892D.6B8CC2E3@trailing-edge.com> Date: Tue, 15 Aug 2000 18:17:17 -0400 From: Tim Shoppa Organization: Trailing Edge Technology X-Mailer: Mozilla 3.03Gold (X11; I; OpenVMS V7.2 AlphaServer 1200 5/533 4MB) MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 28 NNTP-Posting-Host: 63.73.218.130 X-Trace: reader0.news.uu.net 966377838 25089 63.73.218.130 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!newsfeed.mathworks.com!news.mathworks.com!uunet!nyc.uu.net!ffx.uu.net!spool0.news.uu.net!reader0.news.uu.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61998 alt.sys.pdp8:28 alt.sys.pdp10:70 alt.sys.pdp11:22 Max Jester wrote: > > In alt.sys.pdp10 Eric Smith wrote: > > > The only place I remember seeing a 2900 in a PDP-11 was the floating point > > board for the 11/44. In most [*] of the TTL implementations, the CPU > > data path used 74181 or 74S181 ALU chips. > > Um, are you sure about that? I recall when we popped the FP board into > an 11/44 and ISTR that the board consisted of 16 74181s plus TTL glue. I just looked, and the 11/44 FPU's in use here have 16 2901's. The 2901 is a 40-pin package, and the 74181 is a 24-pin package, so I don't think someone's been substituting parts on me :-). (I'll have to tell the story of the 8087 in the FPJ11 socket sometime!) The 74181 is commonly used elsewhere in PDP-11's and peripherals, though. Was there ever an architecture where opcode bits mapped *directly* into 74181 control inputs? A lot of the 74181 control inputs produced rather silly results, so this would seem to me to be a waste of opcode bits at a time when memory was very expensive, but if someone wants to point out a successful or unsuccesful architecture that pulled this trick, I'd appreciate learning about it! Tim. ###### From: torek@elf.bsdi.com (Chris Torek) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Organization: none of the above Lines: 20 Message-ID: <8ncfr6$49s$1@elf.bsdi.com> References: <8n9n30$1tsg$1@news.enteract.com> <8nca8j$3vt$1@elf.bsdi.com> <8ncc95$1bn2$1@news.enteract.com> Date: Tue, 15 Aug 2000 22:22:47 GMT NNTP-Posting-Host: 63.193.109.34 X-Complaints-To: abuse@flash.net X-Trace: news.flash.net 966378167 63.193.109.34 (Tue, 15 Aug 2000 17:22:47 CDT) NNTP-Posting-Date: Tue, 15 Aug 2000 17:22:47 CDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!nntp-out.monmouth.com!newspeer.monmouth.com!howland.erols.net!nntp.flash.net!news.flash.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61971 >Chris Torek wrote: >> The video circuit in the TRS-80 [model I was only 7 bits wide ...] >> The character generator inside the (or at least "my") TRS-80 Model I >> did have a full set of lowercase characters. In article <8ncc95$1bn2$1@news.enteract.com> Eric Fischer wrote: >Didn't they still manage to squeeze a few extra cents out of the >character generator, by choosing one with such badly designed >lowercase letters that no one who actually wanted to use lowercase >would choose it? It is certainly possible that Tandy did such a thing. Mine had no major problems with its lowercase, except that it looked exceedingly odd without descenders. I found myself wishing for a DECwriter font. :-) -- In-Real-Life: Chris Torek, Berkeley Software Design Inc El Cerrito, CA, USA Domain: torek@bsdi.com +1 510 234 3167 http://claw.bsdi.com/torek/ (not always up) I report spam to abuse@. ###### From: jepler.lnk@lnk.ispi.net (jepler epler) Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? References: <8n8c13$1rk$1@pegasus.csx.cam.ac.uk> <8n9432$l66$1@pegasus.csx.cam.ac.uk> <399815F8.2BCA625F@etnus.com> Message-ID: User-Agent: slrn/0.9.5.7 (UNIX) Lines: 17 Date: Wed, 16 Aug 2000 12:42:52 GMT NNTP-Posting-Host: 206.131.194.27 X-Trace: news.randori.com 966429772 206.131.194.27 (Wed, 16 Aug 2000 05:42:52 PDT) NNTP-Posting-Date: Wed, 16 Aug 2000 05:42:52 PDT 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-was.dfn.de!nntp-out.monmouth.com!newspeer.monmouth.com!korova.insync.net!news.randori.com!jepler.lnk Xref: chonsp.franklin.ch alt.folklore.computers:62021 On Tue, 15 Aug 2000 20:17:16 GMT, Michael Wojcik wrote: >Or an AS/400, where the firmware invalidates a pointer that's >manipulated illegally (eg. by writing into it with memcpy) and where >there is no C integer type that's large enough to contain the data in >a C pointer type. In that case, how do you implment C at all? void f(void) { int x, *a=&x, *b; memcpy(b, a, sizeof(int*)); *b = 1; assert(a==b); } Jeff ###### Message-ID: <399A9A69.4EC5D001@mikron.de> Date: Wed, 16 Aug 2000 15:43:05 +0200 From: Bernd Paysan X-Mailer: Mozilla 4.72 [de] (X11; I; Linux 2.2.14 i686) X-Accept-Language: de, en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <399815F8.2BCA625F@etnus.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 34 Organization: Customer of UUNET Deutschland GmbH NNTP-Posting-Host: 194.139.17.38 X-Trace: businessnews.de.uu.net 966434410 11051 194.139.17.38 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!newsfeed00.sul.t-online.de!newsfeed01.sul.t-online.de!t-online.de!do.de.uu.net!nr-do2.de.uu.net!businessnews.de.uu.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62027 Caveman wrote: > > In article , > Jay Maynard wrote: > > > >I have absolutely no idea why C programmers constantly assume that pointers > >and integers are interchangeable, but there's an obscene amount of code out > >there that does exactly that. Don't believe me? Get an Alpha. > > In my experience, the most common reason has always been that few C > programmers seem to really understand what a pointer is. > > This often leads them to substitute integer arithmetic, in which they > have a higher confidence level, though still little real understanding > of how this is implemented at runtime. > > Programmers, in other words, show a certain fear and prejudice toward > pointer ops because they seem mysterious and strange. > > It might be fruitful to develop a programmer psychologist who could > guide the psychological aspects of programmer mindset, just so long > as it's not a dancing paperclip. ;) The main reason I use integer for pointer arithmetic is that pointer arithmetics doesn't provide functions like &, |, or ^. These operations allow to do some tricks like producing cache-line aligned pointers (e.g. (malloc(sizeof(what you need)+sizeof(cacheline)-1)+sizeof(cacheline)-1) & -sizeof(cacheline)). C99 provides intptr_t, which solves the problem whether a pointer is int or long in this or that implementation. -- Bernd Paysan "If you want it done right, you have to do it yourself" http://www.jwdt.com/~paysan/ ###### X-Newsreader: xrn 9.00 From: michael.wojcik@merant.com (Michael Wojcik) Reply-To: michael.wojcik@merant.com Subject: Re: Q: Why not (2^n)-bit? Newsgroups: alt.folklore.computers References: <8n8c13$1rk$1@pegasus.csx.cam.ac.uk> <8n9432$l66$1@pegasus.csx.cam.ac.uk> <399815F8.2BCA625F@etnus.com> Organization: MERANT Inc. Originator: mww@lorelei.michaelwojcik.org Lines: 91 Message-ID: Date: Wed, 16 Aug 2000 19:14:47 GMT NNTP-Posting-Host: 158.252.172.198 X-Complaints-To: abuse@earthlink.net X-Trace: newsread1.prod.itd.earthlink.net 966453287 158.252.172.198 (Wed, 16 Aug 2000 12:14:47 PDT) NNTP-Posting-Date: Wed, 16 Aug 2000 12:14:47 PDT 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!RRZ.Uni-Koeln.DE!news.netcologne.de!newsfeed01.sul.t-online.de!newsfeed00.sul.t-online.de!t-online.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!cpk-news-hub1.bbnplanet.com!news.gtei.net!news.mindspring.net!newsfeed2.earthlink.net!newsfeed.earthlink.net!newsmaster1.prod.itd.earthlink.net!newsread1.prod.itd.earthlink.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:61987 In article , jepler.lnk@lnk.ispi.net (jepler epler) writes: > On Tue, 15 Aug 2000 20:17:16 GMT, Michael Wojcik > wrote: > >Or an AS/400, where the firmware invalidates a pointer that's > >manipulated illegally (eg. by writing into it with memcpy) and where > >there is no C integer type that's large enough to contain the data in > >a C pointer type. > In that case, how do you implment C at all? You write a compiler and execution environment that implements the rules specified for the virtual machine in ISO 9899, just as you would on any other platform. > void f(void) { > int x, *a=&x, *b; > memcpy(b, a, sizeof(int*)); > *b = 1; > assert(a==b); > } What's the problem? (Well, you haven't included the required headers, so memcpy isn't going to be prototyped correctly, which is going to break things. But assuming you had...) a and b are of type "pointer to int". In the AS/400's C implementation, that happens to be a 16-byte structure, but that's opaque to the programmer. The memcpy statement copies the value of x (which is undefined, since X was not initialized) into the region pointed to by b. b is also undefined, so you'll get a machine check right there. You're also trying to copy 16 bytes (thanks to "sizeof(int*)"), so you'd blow up on source and destination buffer overruns even if b were valid. Assuming you had b correctly pointing to a space of the appropriate size and alignment, eg. int x, y, *a=&x, *b=&y; memcpy(b, a, sizeof(int)); Your memcpy would have been successful. However, b and a are still different, so your assert would have failed. Now, let's assume what you meant was something like int x, *a=&x, *b; memcpy(&b, &a, sizeof(int*)); ie. copy the value of a to b, so that b points to the same area that a does. As I interpret ISO 9899 (I'm actually looking at 9899-1990, and the draft of 9899-1999, n869), there are only three ways to assign a value to a pointer variable: - simple assignment (a = &x) - structure assignment - use of the %p conversion specifier in one of the scanf functions, reading data written using a %p specifier with one of the printf functions The standard apparently doesn't require copying a pointer value with memcpy to produce a valid pointer value in the destination. It's nasal-demonic. Now, it may be that AS/400 C allows this; I have some vague memory that some "indirect" pointer assignments work, because the firmware verifies that the source is a valid "space pointer" (the type used by the implementation for the object representation of C pointers). I'd test it, but my AS/400 is down at the moment. But I don't believe it's required to permit it. (And another disclaimer: The AS/400's gone through at least three C implementations, and most of what I know of the internals come from earlier rather than more recent ones. Some of this stuff may have changed.) -- Michael Wojcik michael.wojcik@merant.com AAI Development, MERANT (block capitals are a company mandate) Department of English, Miami University Memory, I realize, can be an unreliable thing; often it is heavily coloured by the circumstances in which one remembers, and no doubt this applies to certain of the recollections I have gathered here. -- Kazuo Ishiguro ###### From: "Charlie Gibbs" Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: 16 Aug 00 11:50:10 -0800 Organization: http://extra.newsguy.com Lines: 13 Message-ID: <410.263T540T7103969@sky.bus.com> References: <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8n91sd$107g$1@ausnews.austin.ibm.com> <39985A75.485B@hda.hydro.com> NNTP-Posting-Host: p-097.newsdawg.com X-Newsreader: THOR 2.5a (Amiga;TCP/IP) 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!rz.uni-karlsruhe.de!schlund.de!newsfeed01.sul.t-online.de!t-online.de!unlisys!news.snafu.de!newsfeedZ.netscum.dQ!netscum.int!pln-e!spln!dex!extra.newsguy.com!newsp.newsguy.com!news1 Xref: chonsp.franklin.ch alt.folklore.computers:62032 In article <39985A75.485B@hda.hydro.com> Terje.Mathisen@hda.hydro.com (Terje Mathisen) writes: >"What, only 2e9 different characters to choose from? That's less than >one per human being!" Then it probably doesn't include the glyph once used by The Artist Formerly Known As "The Artist Formerly Known As 'Prince'". -- cgibbs@sky.bus.com (Charlie Gibbs) Remove the first period after the "at" sign to reply. ###### From: Brian Inglis Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 16 Aug 2000 22:35:21 -0600 Organization: Systematic Software Reply-To: Brian.dot.Inglis@SystematicSw.ab.ca Message-ID: References: <399815F8.2BCA625F@etnus.com> <399A9A69.4EC5D001@mikron.de> X-Newsreader: Forte Agent 1.8/32.548 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit NNTP-Posting-Host: 207.148.137.49 X-Original-NNTP-Posting-Host: 207.148.137.49 X-Trace: 16 Aug 2000 22:35:22 -0700, 207.148.137.49 Lines: 42 X-Original-NNTP-Posting-Host: 204.50.1.43 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!sunqbc.risq.qc.ca!newsfeed.mathworks.com!news.idt.net!nntp.cadvision.com!news.cadvision.com!207.148.137.49 Xref: chonsp.franklin.ch alt.folklore.computers:62078 On Wed, 16 Aug 2000 15:43:05 +0200, Bernd Paysan wrote: >Caveman wrote: >> >> In article , >> Jay Maynard wrote: >> > >> >I have absolutely no idea why C programmers constantly assume that pointers >> >and integers are interchangeable, but there's an obscene amount of code out >> >there that does exactly that. Don't believe me? Get an Alpha. >> >> In my experience, the most common reason has always been that few C >> programmers seem to really understand what a pointer is. >> >> This often leads them to substitute integer arithmetic, in which they >> have a higher confidence level, though still little real understanding >> of how this is implemented at runtime. >> >> Programmers, in other words, show a certain fear and prejudice toward >> pointer ops because they seem mysterious and strange. >> >> It might be fruitful to develop a programmer psychologist who could >> guide the psychological aspects of programmer mindset, just so long >> as it's not a dancing paperclip. ;) > >The main reason I use integer for pointer arithmetic is that pointer >arithmetics doesn't provide functions like &, |, or ^. These operations >allow to do some tricks like producing cache-line aligned pointers (e.g. >(malloc(sizeof(what you need)+sizeof(cacheline)-1)+sizeof(cacheline)-1) >& -sizeof(cacheline)). C99 provides intptr_t, which solves the problem >whether a pointer is int or long in this or that implementation. I believe that the wording may allow intptr_t to be defined if there is an integral type which can hold a pointer -- which appears not to be the case in this implementation -- which would make your code not strictly conforming. Thanks. Take care, Brian Inglis Calgary, Alberta, Canada -- Brian_Inglis@CSi.com (Brian dot Inglis at SystematicSw dot ab dot ca) use address above to reply ###### From: Peter Lindgren Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Thu, 17 Aug 2000 10:23:58 +0200 Organization: Ericsson Erisoft AB Lines: 44 Message-ID: <399BA11E.BA1A907B@epl.ericsson.se> References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8nf2i9$2hgk$1@nntp1.ba.best.com> NNTP-Posting-Host: lws300.lu.erisoft.se Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.7 [en] (X11; I; SunOS 5.6 sun4u) X-Accept-Language: Swedish, sv, en Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!uni-erlangen.de!newsfeed1.telenordia.se!algonet!uab.ericsson.se!erinews.ericsson.se!canaio.lu.erisoft.se!antares.lu.erisoft.se!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62044 Joe Smith wrote: > > In article <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com>, > dls2 wrote: > >"IBM PC Extended ASCII Display Characters > > Strictly speaking, the ASCII character set only includes values > >up to 127 decimal (7F hex). However, when the IBM PC was > >developed, the video card contained one byte for each character > >in the 80x25 character display. Gee...what to do with that extra bit > >per character? Why not invent 128 new characters, for line-drawing > >[snip] http://www.jimprice.com/jim-asc.htm > > > >If this is true, then why did the video card contain one (8-bit) byte > >for each character? > > However, the video card actually used two (8-bit) bytes for each character > position on the display. One 8-bit quantity to select one of 256 glyphs, > and two 4-bit quantities (3+1 and 3+1) to select the foreground and > background colors. Sorry, but the initial IBM PC didn't have colors. It was the MDA, Monochrome Display Adapter, to which Hercules added "Hercules graphics", a monochrome graphic mode (only one intensity) I think in the range 720x348 pixels. IIRC, the MDA had 8-bit characters and an 8-bit attribute interleaved in the on-board memory. Or was that Hercules addition? The Hercules card at least, you could program it to interpret one of the attribute bits as "flashing foreground" or "high intensive background". Other attributes were "underline" and "background on/off". Aahhh, those were the days of the 6845 graphics controller. It was farely often used, wasn't it? It was used in the Swedish Luxor's computers ABC 80 (I guess) and ABC 800 (I know). I actually have the schematics for my Hercules board around, anyone care for a .jpeg in their mailbox? /Peter -- Peter Lindgren Bachelor of Computer Engineering ERIEYE C2 Software Design http://www.ericsson.se/SE/epl Get your own AWACS: http://www.ericsson.se/microwave/2-Products/erieye.asp --- Opinions above, expressed or implicit, are my personal opinons only --- ###### Message-ID: <399C0472.29D0AD05@mikron.de> Date: Thu, 17 Aug 2000 17:27:46 +0200 From: Bernd Paysan X-Mailer: Mozilla 4.72 [de] (X11; I; Linux 2.2.14 i686) X-Accept-Language: de, en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <8nbbvn$1fks$1@citadel.in.taronga.com> <39987007.4DFFBF23@jetnet.ab.ca> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 23 Organization: Customer of UUNET Deutschland GmbH NNTP-Posting-Host: 194.139.17.38 X-Trace: businessnews.de.uu.net 966528020 12807 194.139.17.38 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!uni-erlangen.de!newsfeed.germany.net!newsfeed01.sul.t-online.de!t-online.de!do.de.uu.net!nr-do2.de.uu.net!businessnews.de.uu.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62085 Ariel Scolnicov wrote: > There is no advantage in making RGB values floating point, since they > belong to a fixed range with a fixed resolution. True, maybe you need > to take a logarithm (or apply a more complicated "transfer function"), > but the range is going to be bounded and finite. Not completely. Physical rendering needs the energy density (in candela) of the light, while on the frame buffer, you "only" need perceived colors. That part is solved "good enough" (with 8:8:8 RGB and gamma=2.2), especially since the BW contrast is also finite. With a physical rendering, you can do quite some nice effects (like clipping the most intensive lights, blurring that image (perhaps with a star-shaped mask), and adding it again - now 100 times darker - to the image, gives nice stars around the lights and reflections). I'd use 16 bit logarithmic representation in the back buffer, and have 16 bit logarithmic add in hardware (that's cheap enough, especially since we can do without multiplication: that's an integer add). -- Bernd Paysan "If you want it done right, you have to do it yourself" http://www.jwdt.com/~paysan/ ###### From: Kevin Strietzel Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Thu, 17 Aug 2000 17:01:14 -0700 Organization: Stratus Computer Lines: 36 Message-ID: <399C7CCA.7F2DE606@stratus.com> References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8nf2i9$2hgk$1@nntp1.ba.best.com> <399BA11E.BA1A907B@epl.ericsson.se> NNTP-Posting-Host: nadaburg.az.stratus.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.7 [en] (WinNT; I) X-Accept-Language: en 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!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!cpk-news-hub1.bbnplanet.com!cambridge1-snf1.gtei.net!news.gtei.net!transfer.stratus.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62107 Peter Lindgren wrote: > > Joe Smith wrote: ... > > However, the video card actually used two (8-bit) bytes for each character > > position on the display. One 8-bit quantity to select one of 256 glyphs, > > and two 4-bit quantities (3+1 and 3+1) to select the foreground and > > background colors. > > Sorry, but the initial IBM PC didn't have colors. It was the MDA, > Monochrome Display Adapter, to which Hercules added "Hercules > graphics", a monochrome graphic mode (only one intensity) I think > in the range 720x348 pixels. > > IIRC, the MDA had 8-bit characters and an 8-bit attribute interleaved > in the on-board memory. Or was that Hercules addition? > > The Hercules card at least, you could program it to interpret one of > the attribute bits as "flashing foreground" or "high intensive > background". Other attributes were "underline" and "background on/off". > Aahhh, those were the days of the 6845 graphics controller. It was > farely often used, wasn't it? It was used in the Swedish Luxor's > computers ABC 80 (I guess) and ABC 800 (I know). You're both right :-). The original IBM PC was announced with two display adapters available: the Monochrome Display and Parallel Printer Adapter, and the Color Graphics Adapter; they only announced the monochrome *display*, and the color display much later. Both worked more or less as described. I don't remember exactly what the attributes were for the MDA, but I remember it having Highlight, Blink, and Underline. Maybe it had independent control of foreground and background? Anyway, --Kevin Strietzel Not speaking for Stratus. ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <39b8a2f5.64050421@news.remarq.com> <8neegc$uqa$1@citadel.in.taronga.com> <399f28b8.6471472@news.remarq.com> <8nfdm3$1g75$1@citadel.in.taronga.com> <399BCB14.5AEDDA11@netinsight.se> <39efde1a.52912321@news.remarq.com> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. X-Newsreader: Gnus v5.5/Emacs 20.3 Date: 17 Aug 2000 17:15:40 -0700 Message-ID: Lines: 10 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 17 Aug 2000 17:21:02 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!unlisys!news.snafu.de!fu-berlin.de!news.maxwell.syr.edu!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62167 alt.sys.pdp8:50 alt.sys.pdp10:100 alt.sys.pdp11:39 ian@beathoven.com (paramucho) writes: > How could you write realtime applications for a machine that didn't > know where the bus would be located on the next model... What on earth does the bus of the "next model" have to do with being able to write a realtime application? I wrote several realtime applications for the IBM PC/AT, and they still run perfectly fine today on machines that don't even *have* a bus, much less an ISA bus. ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <399C161E.783565F@mchpDOTsiemens.de> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. X-Newsreader: Gnus v5.5/Emacs 20.3 Date: 17 Aug 2000 17:27:28 -0700 Message-ID: Lines: 4 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 17 Aug 2000 17:32:50 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!unlisys!news.snafu.de!newsfeed01.sul.t-online.de!newsfeed00.sul.t-online.de!t-online.de!grolier!fr.usenet-edu.net!usenet-edu.net!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62161 alt.sys.pdp8:48 alt.sys.pdp10:96 alt.sys.pdp11:37 Konrad Schwarz writes: > The IBM 360 introduced the 8-bit byte. Actually, the IBM Stretch (AKA IBM 7030 Data Processing System). ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8nf2i9$2hgk$1@nntp1.ba.best.com> <399BA11E.BA1A907B@epl.ericsson.se> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 17 Aug 2000 17:34:31 -0700 Message-ID: Lines: 10 X-Newsreader: Gnus v5.5/Emacs 20.3 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 17 Aug 2000 17:39:53 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!unlisys!news.snafu.de!pr0npusher.netscum.dk!netscum.int!144.212.100.101!newsfeed.mathworks.com!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62164 Peter Lindgren writes: > Sorry, but the initial IBM PC didn't have colors. It was the MDA, > Monochrome Display Adapter, to which Hercules added "Hercules > graphics", a monochrome graphic mode (only one intensity) I think > in the range 720x348 pixels. False. The CGA (Color Graphics Adapter) was introduced on the same day as the MDA and the PC itself. I saw one available for sale in a ComputerLand store on that date. August of 1981, IIRC, but the day escapes me. ###### From: ian@beathoven.com (paramucho) Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? Date: Fri, 18 Aug 2000 01:37:15 GMT Organization: www.beathoven.com Lines: 22 Message-ID: <39a88b65.11279772@news.remarq.com> References: <39b8a2f5.64050421@news.remarq.com> <8neegc$uqa$1@citadel.in.taronga.com> <399f28b8.6471472@news.remarq.com> <8nfdm3$1g75$1@citadel.in.taronga.com> <399BCB14.5AEDDA11@netinsight.se> <39efde1a.52912321@news.remarq.com> Reply-To: ian@beathoven.com X-Complaints-To: newsabuse@supernews.com X-Newsreader: Forte Agent 1.5/32.451 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit 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!sn-xit-03!supernews.com!sn-inject-01!news.supernews.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62123 alt.sys.pdp8:42 alt.sys.pdp10:87 alt.sys.pdp11:32 On 17 Aug 2000 17:15:40 -0700, Eric Smith wrote: >ian@beathoven.com (paramucho) writes: >> How could you write realtime applications for a machine that didn't >> know where the bus would be located on the next model... > >What on earth does the bus of the "next model" have to do with being >able to write a realtime application? I wrote several realtime >applications for the IBM PC/AT, and they still run perfectly fine today >on machines that don't even *have* a bus, much less an ISA bus. The unibus is mapped to a specific physical address range within a VAX. That was where the realtime device registers were. The address range would change from model to model. The applications would not run perfectly fine. Ian ###### From: jmaynard@thebrain.conmicro.cx (Jay Maynard) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 18 Aug 2000 02:04:26 GMT Organization: Neosoft (using Airnews.net!) Lines: 9 Message-ID: <3EF934C3AD9CBA24.2375B83E8FFF0D9B.396FF51A0658B92D@lp.airnews.net> X-Orig-Message-ID: References: <399815F8.2BCA625F@etnus.com> <399A9A69.4EC5D001@mikron.de> Reply-To: jmaynard@conmicro.cx Abuse-Reports-To: abuse at airmail.net to report improper postings NNTP-Proxy-Relay: library2.airnews.net NNTP-Posting-Time: Thu Aug 17 21:04:26 2000 NNTP-Posting-Host: !X;1@1k-Y%@eA&= (Encoded at Airnews!) User-Agent: slrn/0.9.5.4 (UNIX) 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!cpk-news-hub1.bbnplanet.com!news.gtei.net!howland.erols.net!news-out.nntp.airnews.net.MISMATCH!cabal10.airnews.net!news.airnews.net!cabal14.airnews.net!news.airnews.net!cabal2.airnews.net!news-f.iadfw.net!jmaynard Xref: chonsp.franklin.ch alt.folklore.computers:62118 On Wed, 16 Aug 2000 15:43:05 +0200, Bernd Paysan wrote: >The main reason I use integer for pointer arithmetic is that pointer >arithmetics doesn't provide functions like &, |, or ^. These operations >allow to do some tricks like producing cache-line aligned pointers (e.g. >(malloc(sizeof(what you need)+sizeof(cacheline)-1)+sizeof(cacheline)-1) >& -sizeof(cacheline)). C99 provides intptr_t, which solves the problem >whether a pointer is int or long in this or that implementation. Just what machine do you think all the world is, then? ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <39b8a2f5.64050421@news.remarq.com> <8neegc$uqa$1@citadel.in.taronga.com> <399f28b8.6471472@news.remarq.com> <8nfdm3$1g75$1@citadel.in.taronga.com> <399BCB14.5AEDDA11@netinsight.se> <39efde1a.52912321@news.remarq.com> <39a88b65.11279772@news.remarq.com> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 17 Aug 2000 19:46:43 -0700 Message-ID: Lines: 11 X-Newsreader: Gnus v5.5/Emacs 20.3 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 17 Aug 2000 19:52:06 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62163 alt.sys.pdp8:49 alt.sys.pdp10:97 alt.sys.pdp11:38 ian@beathoven.com (paramucho) writes: > The unibus is mapped to a specific physical address range within a > VAX. That was where the realtime device registers were. The address > range would change from model to model. The applications would not run > perfectly fine. So you change a few numbers in a configuration file, recompile, and it works. Since when did anyone ever *really* expect to upgrade hardware with NO impact on software? ###### From: ian@beathoven.com (paramucho) Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? Date: Fri, 18 Aug 2000 03:11:52 GMT Organization: www.beathoven.com Lines: 19 Message-ID: <39aaa96b.18967825@news.remarq.com> References: <39b8a2f5.64050421@news.remarq.com> <8neegc$uqa$1@citadel.in.taronga.com> <399f28b8.6471472@news.remarq.com> <8nfdm3$1g75$1@citadel.in.taronga.com> <399BCB14.5AEDDA11@netinsight.se> <39efde1a.52912321@news.remarq.com> <39a88b65.11279772@news.remarq.com> Reply-To: ian@beathoven.com X-Complaints-To: newsabuse@supernews.com X-Newsreader: Forte Agent 1.5/32.451 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!sn-xit-03!supernews.com!sn-inject-01!news.supernews.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62133 alt.sys.pdp8:43 alt.sys.pdp10:88 alt.sys.pdp11:33 On 17 Aug 2000 19:46:43 -0700, Eric Smith wrote: >ian@beathoven.com (paramucho) writes: >> The unibus is mapped to a specific physical address range within a >> VAX. That was where the realtime device registers were. The address >> range would change from model to model. The applications would not run >> perfectly fine. > >So you change a few numbers in a configuration file, recompile, and it >works. > >Since when did anyone ever *really* expect to upgrade hardware with >NO impact on software? Since one had a PDP-11. Ian ###### From: cecchi@signa.rchland.ibm.com (Del Cecchi) Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? Date: 18 Aug 2000 12:53:21 GMT Organization: IBM Corp. Lines: 20 Message-ID: <8njbk1$r7k$1@news.rchland.ibm.com> References: <39b8a2f5.64050421@news.remarq.com> <8neegc$uqa$1@citadel.in.taronga.com> <399f28b8.6471472@news.remarq.com> <8nfdm3$1g75$1@citadel.in.taronga.com> <399BCB14.5AEDDA11@netinsight.se> <39efde1a.52912321@news.remarq.com> <39a88b65.11279772@news.remarq.com> Reply-To: dcecchi@vnet.ibm.com NNTP-Posting-Host: signa.rchland.ibm.com X-Trace: news.rchland.ibm.com 966603201 27892 9.5.54.183 (18 Aug 2000 12:53:21 GMT) X-Complaints-To: usenet@news.rchland.ibm.com NNTP-Posting-Date: 18 Aug 2000 12:53:21 GMT X-Newsreader: xrn 9.01-beta-3 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.ision.net!ision!bignews.mediaways.net!abq.news.ans.net!news.chips.ibm.com!newsfeed.btv.ibm.com!news!signa.rchland.ibm.com!cecchi Xref: chonsp.franklin.ch alt.folklore.computers:62150 alt.sys.pdp8:46 alt.sys.pdp10:91 alt.sys.pdp11:35 In article , Eric Smith writes: |> ian@beathoven.com (paramucho) writes: |> > The unibus is mapped to a specific physical address range within a |> > VAX. That was where the realtime device registers were. The address |> > range would change from model to model. The applications would not run |> > perfectly fine. |> |> So you change a few numbers in a configuration file, recompile, and it |> works. |> |> Since when did anyone ever *really* expect to upgrade hardware with |> NO impact on software? Since they bought an AS400. :-) -- Del Cecchi cecchi@rchland ###### Message-ID: <399D7571.8B89E801@cmc.com> Date: Fri, 18 Aug 2000 10:42:09 -0700 From: Lars Poulsen X-Mailer: Mozilla 4.73 [en] (Win98; I) X-Accept-Language: da,en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,alt.sys.pdp11,vmsnet.pdp-11 To: Peter da Silva Subject: Re: Q: Why not (2^n)-bit? References: <8ni60u$2v86$1@citadel.in.taronga.com> <399D01B3.1FD28DFE@netinsight.se> <8njm6r$ni9$1@citadel.in.taronga.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 23 NNTP-Posting-Host: 207.154.90.88 X-Trace: azure.impulse.net 966620478 195 207.154.90.88 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!informatik.tu-muenchen.de!news.informatik.uni-muenchen.de!uni-erlangen.de!newsfeeds.belnet.be!news.belnet.be!skynet.be!nntp-relay.ihug.net!ihug.co.nz!sienna.impulse.net!azure.impulse.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62177 alt.sys.pdp11:40 Peter da Silva wrote: > According to my 1969 handbook, the only difference between the /10 and the /20 > was configuration. If that's not the case, and they reused the /10 designation > later on, then in all my previous comments about /10 and /20 elide the /10. The 11/10 that was described in your quote from the '69 handbook either was never built, or it may have been called the 11/15 when it shipped. The 11/10 that did ship (and which I have used) was a micro- programmed re-implementation of the 11/20's capabilities in a half-sized package. It was available in a 5.25" high box with the backplane mounted vertically so the boards were horizontal, or in the 10.5" (BA11-K ?) box with the backplane on the bottom, and the cards mounted vertically. The 11/05 was the name of the 11/10 when it was sold through the OEM channel. (Group list trimmed!) -- / Lars Poulsen - http://www.cmc.com/lars - lars@cmc.com 125 South Ontare Road, Santa Barbara, CA 93105 - +1-805-569-5277 ###### Sender: meissner@tiktok.cygnus.com Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <39b8a2f5.64050421@news.remarq.com> <8neegc$uqa$1@citadel.in.taronga.com> <399f28b8.6471472@news.remarq.com> <8nfdm3$1g75$1@citadel.in.taronga.com> <399BCB14.5AEDDA11@netinsight.se> <39efde1a.52912321@news.remarq.com> <39a88b65.11279772@news.remarq.com> <39aaa96b.18967825@news.remarq.com> From: Michael Meissner Message-ID: Organization: Red Hat, Inc.. Lines: 37 X-Newsreader: Gnus v5.7/Emacs 20.7 Date: 18 Aug 2000 14:28:15 -0400 NNTP-Posting-Host: 209.192.197.34 X-Complaints-To: abuse@shore.net X-Trace: news.shore.net 966623301 209.192.197.34 (Fri, 18 Aug 2000 14:28:21 EDT) NNTP-Posting-Date: Fri, 18 Aug 2000 14:28:21 EDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!sunqbc.risq.qc.ca!newsfeed.mathworks.com!news.shore.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62156 alt.sys.pdp8:47 alt.sys.pdp10:93 alt.sys.pdp11:36 ian@beathoven.com (paramucho) writes: > On 17 Aug 2000 19:46:43 -0700, Eric Smith > wrote: > > >ian@beathoven.com (paramucho) writes: > >> The unibus is mapped to a specific physical address range within a > >> VAX. That was where the realtime device registers were. The address > >> range would change from model to model. The applications would not run > >> perfectly fine. > > > >So you change a few numbers in a configuration file, recompile, and it > >works. > > > >Since when did anyone ever *really* expect to upgrade hardware with > >NO impact on software? > > Since one had a PDP-11. Outside of the user level integer instructions, the PDP family was horrible about inter-line compatibiliy. The supervisor instructions were different, and there a couple of different floating point units that used different instructions. For instance if I recall the details correctly, if you have a PDP-11/45, a PDP-11/40, and an LSI-11/02, and you want the exact same code to run on all machines, and want to set the interrupt priority level, how do you do it? Each of the machines has a different instruction to set the interrupt priority. Finally, I just created a fake interrupt stack and did a return from interrupt instruction, which was the only non-model specific method I could figure out to set the priority. -- Michael Meissner, Red Hat, Inc. PMB 198, 174 Littleton Road #3, Westford, Massachusetts 01886, USA Work: meissner@redhat.com phone: +1 978-486-9304 Non-work: meissner@spectacle-pond.org fax: +1 978-692-4482 ###### From: peter@nmti.com (Peter da Silva) Newsgroups: alt.folklore.computers,alt.sys.pdp11,vmsnet.pdp-11 Subject: Re: Q: Why not (2^n)-bit? Date: 18 Aug 2000 18:41:37 GMT Organization: Engineering Netwrk Support, ABBNM Lines: 22 Message-ID: <8nk011$ngq@web.nmti.com> References: <399D01B3.1FD28DFE@netinsight.se> <8njm6r$ni9$1@citadel.in.taronga.com> <399D7571.8B89E801@cmc.com> NNTP-Posting-Host: grendel.eng.baileynm.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!news-stu1.dfn.de!news-koe1.dfn.de!news-fra1.dfn.de!news0.de.colt.net!colt.net!diablo.theplanet.net!news.maxwell.syr.edu!newsfeed.stanford.edu!news-proxy.baileynm.com!web.nmti.com!peter Xref: chonsp.franklin.ch alt.folklore.computers:62097 alt.sys.pdp11:30 In article <399D7571.8B89E801@cmc.com>, Lars Poulsen wrote: > The 11/10 that did ship (and which I have used) was a micro- > programmed re-implementation of the 11/20's capabilities in > a half-sized package. It was available in a 5.25" high box > with the backplane mounted vertically so the boards were > horizontal, or in the 10.5" (BA11-K ?) box with the backplane on > the bottom, and the cards mounted vertically. > > The 11/05 was the name of the 11/10 when it was sold through > the OEM channel. Ah, that explains it then. At Berkeley the assembly language course was taught in "11/10" boxes in the 10.5" cabinet, and "11/05" boxes in the 5.25" box. Since the 11/10 I used was the size of the 11/20 CPU, and it was over 20 years ago... -- `-_-' In hoc signo hack, Peter da Silva. 'U` "Milloin halasit viimeksi suttasi?" Disclaimer: WWFD? ###### Message-ID: <399E6477.398903A4@trailing-edge.com> Date: Sat, 19 Aug 2000 10:41:59 -0400 From: Tim Shoppa Organization: Trailing Edge Technology X-Mailer: Mozilla 3.03Gold (X11; I; OpenVMS V7.2 AlphaServer 1200 5/533 4MB) MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <39b8a2f5.64050421@news.remarq.com> <8neegc$uqa$1@citadel.in.taronga.com> <399f28b8.6471472@news.remarq.com> <8nfdm3$1g75$1@citadel.in.taronga.com> <399BCB14.5AEDDA11@netinsight.se> <39efde1a.52912321@news.remarq.com> <39a88b65.11279772@news.remarq.com> <39aaa96b.18967825@news.remarq.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 41 NNTP-Posting-Host: 63.73.218.130 X-Trace: reader1.news.uu.net 966696120 13980 63.73.218.130 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!feed2.onemain.com!feed1.onemain.com!uunet!ffx.uu.net!spool1.news.uu.net!spool0.news.uu.net!reader1.news.uu.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62134 alt.sys.pdp8:44 alt.sys.pdp10:89 alt.sys.pdp11:34 paramucho wrote: > > On 17 Aug 2000 19:46:43 -0700, Eric Smith > wrote: > > >ian@beathoven.com (paramucho) writes: > >> The unibus is mapped to a specific physical address range within a > >> VAX. That was where the realtime device registers were. The address > >> range would change from model to model. The applications would not run > >> perfectly fine. > > > >So you change a few numbers in a configuration file, recompile, and it > >works. > > > >Since when did anyone ever *really* expect to upgrade hardware with > >NO impact on software? > > Since one had a PDP-11. We're lucky that all the "real" PDP-11's have such a consistent bus model. Whether a PDP-11/20 or an 11/03 or a 11/70 or a 11/94, things are pretty much the same to the code. (OK, there are some inconsistencies when it comes to 18 vs 22 bit addressing, and interrupt priorities could get a little strange, but these were "straightforward" considerations.) And there was always the "system console" at the same place, whether it really be a hardcopy terminal or a CRT or another computer. Now the desktop PDP-11's (i.e. DEC PRO 350 and 380) and the single-board PDP-11's (the Falcon, various non-DEC products based around the KDF11 and KDJ11) are a different story. Some of these don't even have anything resembling a "console device". VAXen have had much more variation in their buses (everything from real Q-bus and Unibusses to busless/dickless machines to Turbochannel and BI-bus and XMI) and their console systems have shown great variation - from 11/03's to PRO 350's to 8085's and some that aren't even describable without drawing pictures :-). Tim.  ###### From: fixxit@bright.net Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Message-ID: <40kupssgofs79lv592b74nsk8scrtvo1j5@4ax.com> References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8nf2i9$2hgk$1@nntp1.ba.best.com> <399BA11E.BA1A907B@epl.ericsson.se> X-Newsreader: Forte Agent 1.8/16.548 X-No-Archive: yes MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 29 Date: Sun, 20 Aug 2000 23:06:31 -0400 NNTP-Posting-Host: 216.201.28.166 X-Complaints-To: abuse@bright.net X-Trace: cletus.bright.net 966827037 216.201.28.166 (Sun, 20 Aug 2000 23:03:57 EDT) NNTP-Posting-Date: Sun, 20 Aug 2000 23:03:57 EDT Organization: bright.net Ohio 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!howland.erols.net!newsfeed.skycache.com!Cidera!cletus.bright.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62260 Eric Smith wrote: >Peter Lindgren writes: >> Sorry, but the initial IBM PC didn't have colors. It was the MDA, >> Monochrome Display Adapter, to which Hercules added "Hercules >> graphics", a monochrome graphic mode (only one intensity) I think >> in the range 720x348 pixels. > >False. The CGA (Color Graphics Adapter) was introduced on the same day >as the MDA and the PC itself. I saw one available for sale in a >ComputerLand store on that date. August of 1981, IIRC, but the day >escapes me. Um, I think he said "announced" as opposed to "introduced", in the part of the thread that got snipped. The "initial IBM PC" was introduced August 12, 1981. It could be got with max 256K of RAM on the motherboard, and 1 or 2 double density 5 1/4" drives (720k 3 1/2" was optional). It didn't even come with a hard drive, altho a "PC expansion unit" with a whopping 10M hard drive could be had. My book doesn't mention video (_Upgrading and Repairing PCs_), but I'm pretty sure CGA was available as an option with the original. When replying, send a copy via email--sometimes I don't keep up with the group! Note the URL for email corrections--Visit my web page to see what I do... www.bright.net/~fixit ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? References: <39a013e5.25745326@news.remarq.com> <399941CC.1CC6DFA5@netinsight.se> <8noeok$q6t$6@bob.news.rcn.net> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. X-Newsreader: Gnus v5.5/Emacs 20.3 Date: 20 Aug 2000 20:24:28 -0700 Message-ID: Lines: 23 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 20 Aug 2000 20:30:25 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62254 alt.sys.pdp8:76 alt.sys.pdp10:138 alt.sys.pdp11:47 jmfbahciv@aol.com writes: > Now I'm not sure what you mean by architecture. Don't you > consider the KA and the KL, both PDP-10s, different hardware > architecture? We sure did. In the Blaauw & Brooks taxonomy [*], architecture is what is seen by the programmers. In that sense, the PDP-6 and KA10 are nearly identical. The later PDP-10 processors are mostly the same, with the addition of paging, which is mostly only an issue for the monitor developers. The implementation is how the logic design implements the architecture. The KA, KI, KL, and KS had very different implementations. Even within an implementation, there can be large differences in the realization, which is how the logic is physically implemented. For instance, the IBM 709 and 7090 had the same architecture and implementation, but the realization was tubes for the 709 and transistors for the 7090. [*] _Computer Architecture: Concepts and Evolution_, Gerrit A> Blaauw & Frederick P. Brooks Jr, Addison-Wesley, 1997 http://www.amazon.com/exec/obidos/ASIN/0201105578/ ###### From: jmfbahciv@aol.com Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? Date: Tue, 22 Aug 00 09:57:22 GMT Organization: UltraNet Communications, Inc. Lines: 43 Message-ID: <8nttge$eis$11@bob.news.rcn.net> References: <39a013e5.25745326@news.remarq.com> <399941CC.1CC6DFA5@netinsight.se> <8noeok$q6t$6@bob.news.rcn.net> X-Trace: rRC5BYIPRpVf+o57LtGtfdUZdk4Kj+zwNBRsLng2e0o= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 22 Aug 2000 12:59:58 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!newsfeeds.belnet.be!news.belnet.be!feed2.onemain.com!feed1.onemain.com!dc1.nntp.concentric.net!newsfeed.concentric.net!feed1.news.rcn.net!rcn!207-172-245-83 Xref: chonsp.franklin.ch alt.folklore.computers:62330 alt.sys.pdp8:89 alt.sys.pdp10:167 alt.sys.pdp11:58 In article , Eric Smith wrote: >jmfbahciv@aol.com writes: >> Now I'm not sure what you mean by architecture. Don't you >> consider the KA and the KL, both PDP-10s, different hardware >> architecture? We sure did. > >In the Blaauw & Brooks taxonomy [*], architecture is what is seen by the >programmers. In that sense, the PDP-6 and KA10 are nearly identical. >The later PDP-10 processors are mostly the same, with the addition >of paging, which is mostly only an issue for the monitor developers. Not at all. There were more differences than paging. One of design goals of each version of the processor was backwards compatibility. I don't think that necessarily confines a new architecture. I can't think of a single programming level that wasn't affected by each CPU implementation. I suppose it was mostly invisible at the compiler level in the cases of FORTRAN and COBOL. But I can think of aspects that even that level of programming changed, especially with advent of extended addressing on the KL. > >The implementation is how the logic design implements the architecture. >The KA, KI, KL, and KS had very different implementations. > >Even within an implementation, there can be large differences in the >realization, which is how the logic is physically implemented. For >instance, the IBM 709 and 7090 had the same architecture and implementation, >but the realization was tubes for the 709 and transistors for the 7090. > > > >[*] _Computer Architecture: Concepts and Evolution_, >Gerrit A> Blaauw & Frederick P. Brooks Jr, Addison-Wesley, 1997 >http://www.amazon.com/exec/obidos/ASIN/0201105578/ I wasn't considering the electronic components as a litmus test. /BAH Subtract a hundred and four for e-mail. ###### From: handleym@ricochet.net (Maynard Handley) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Mon, 21 Aug 2000 12:22:56 -0700 Organization: None Lines: 25 Message-ID: References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> NNTP-Posting-Host: handma2.apple.com X-Trace: news.apple.com 966885762 5452 17.202.32.152 (21 Aug 2000 19:22:42 GMT) X-Complaints-To: usenet@news.apple.com NNTP-Posting-Date: 21 Aug 2000 19:22:42 GMT X-Newsreader: MT-NewsWatcher 2.3.1 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!unlisys!news.snafu.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.stanford.edu!paloalto-snf1.gtei.net!news.gtei.net!forum.apple.com!news.apple.com!handma2.apple.com!user Xref: chonsp.franklin.ch alt.folklore.computers:62357 In article <399D9B1D.1E0CE346@dusknet.dhs.org>, Jeff Teunissen wrote: >Nick Maclaren wrote: >> >> In article <39985C56.27AF@hda.hydro.com>, >> Terje Mathisen wrote: >[snip] >> >John Carmack wrote a long (.plan?) message this spring, calling for >> >64-bit graphics, using 16 bits (in fp format) for each of R, G & B, plus >> >another 16 bits for Alpha. >> >> Hmm. What species does he belong to? Mere humans cannot distinguish >> more than about 2^20 colour/intensity combinations .... > >It's an OpenGL thing. When you do things like applying shading, glow maps, >and so on to a base texture (particularly with variable alpha) in "only" >32bpp, banding is a common result. Doing it in 64 bits and sampling down to >32 (which is what Carmack's .plan was talking about) helps eliminate the >problem. It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp in 32 bits. Maynard ###### Message-ID: <39A1DC28.A694DCBA@ibm.net> Date: Mon, 21 Aug 2000 20:49:28 -0500 From: Del Cecchi X-Mailer: Mozilla 4.7 [en]C-CCK-MCD NSCPCD47 (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit NNTP-Posting-Host: 32.100.32.121 X-Trace: 22 Aug 2000 01:50:25 GMT, 32.100.32.121 Organization: Global Network Services - Remote Access Mail & News Services Lines: 35 X-Complaints-To: abuse@prserv.net 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!feed2.onemain.com!feed1.onemain.com!europa.netcrusader.net!4.1.16.34!cpk-news-hub1.bbnplanet.com!news.gtei.net!newsfeed.us.ibm.net!ibm.net!news3.prserv.net!32.100.32.121 Xref: chonsp.franklin.ch alt.folklore.computers:62371 Maynard Handley wrote: > > In article <399D9B1D.1E0CE346@dusknet.dhs.org>, Jeff Teunissen > wrote: > > >Nick Maclaren wrote: > >> > >> In article <39985C56.27AF@hda.hydro.com>, > >> Terje Mathisen wrote: > >[snip] > >> >John Carmack wrote a long (.plan?) message this spring, calling for > >> >64-bit graphics, using 16 bits (in fp format) for each of R, G & B, plus > >> >another 16 bits for Alpha. > >> > >> Hmm. What species does he belong to? Mere humans cannot distinguish > >> more than about 2^20 colour/intensity combinations .... > > > >It's an OpenGL thing. When you do things like applying shading, glow maps, > >and so on to a base texture (particularly with variable alpha) in "only" > >32bpp, banding is a common result. Doing it in 64 bits and sampling down to > >32 (which is what Carmack's .plan was talking about) helps eliminate the > >problem. > > It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp > in 32 bits. > > Maynard I don't do graphics, so please bear with me. Do you mean the eye can perceive an intensity difference of 1 in 2**32? Or do you mean something else by grey ramp? Also it seems as if one would have only about 1 tenmillionth of the ramp displayed because the screen is only on the order of 1000 pixels out of 10 billion intensity levels. del cecchi ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 21 Aug 2000 19:07:11 -0700 Message-ID: Lines: 17 X-Newsreader: Gnus v5.5/Emacs 20.3 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 21 Aug 2000 19:13:18 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!sunqbc.risq.qc.ca!newsfeed.mathworks.com!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62361 Maynard Handley wrote: > It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp > in 32 bits. Del Cecchi writes: > I don't do graphics, so please bear with me. Do you mean the eye can > perceive an intensity difference of 1 in 2**32? When discussing "32-bit color", usually one means 8 bits per RGB component. Maynard means that the eye can perceive an intensity difference of 1 in 2**8. Others have claimed that they eye cannot distinguish 2*24 colors, so 8 bits per component is good enough. Maynard points out (correctly, IMNSHO) that it is not. ###### From: krw@attglobal.net (Keith R. Williams) Reply-To: krw@attglobal.net Message-ID: Newsgroups: comp.arch,alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8nf2i9$2hgk$1@nntp1.ba.best.com> <399BA11E.BA1A907B@epl.ericsson.se> <40kupssgofs79lv592b74nsk8scrtvo1j5@4ax.com> X-Newsreader: ProNews/2 Version 1.50 ia093a MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8bit NNTP-Posting-Host: 32.101.168.119 Date: 22 Aug 2000 02:24:54 GMT X-Trace: 22 Aug 2000 02:24:54 GMT, 32.101.168.119 Organization: Global Network Services - Remote Access Mail & News Services Lines: 58 X-Complaints-To: abuse@prserv.net Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!cpk-news-hub1.bbnplanet.com!news.gtei.net!newsfeed.us.ibm.net!ibm.net!news3.prserv.net!32.101.168.119 Xref: chonsp.franklin.ch alt.folklore.computers:62356 On Mon, 21 Aug 2000 03:06:31, fixxit@bright.net wrote: > Eric Smith wrote: > > >Peter Lindgren writes: > >> Sorry, but the initial IBM PC didn't have colors. It was the MDA, > >> Monochrome Display Adapter, to which Hercules added "Hercules > >> graphics", a monochrome graphic mode (only one intensity) I think > >> in the range 720x348 pixels. > > > >False. The CGA (Color Graphics Adapter) was introduced on the same day > >as the MDA and the PC itself. I saw one available for sale in a > >ComputerLand store on that date. August of 1981, IIRC, but the day > >escapes me. > > Um, I think he said "announced" as opposed to "introduced", in the > part of the thread that got snipped. No, it was both announced and available. > The "initial IBM PC" was introduced August 12, 1981. It could be got > with max 256K of RAM on the motherboard, Nope! It had a standard 16KB and options (on the motherboard) for up to 64KB. There were I/O channel cards available with 32KB and 64KB, but they were *expen$ive* (IIRC the 64KB card was $500). You're thinking about the so-called PC-2, that had 64K DRAMs and could be coaxed (with a 74LS139, IIRC) to take 256K chips. The PC-2 came out at the same time as the XT. > and 1 or 2 double density 5 1/4" drives (720k 3 1/2" was optional). ..again wrong. The first PC had one or two double-density, but *single*-sided 40-track, 8-sector/track (that means 140K per diskette) drives. > It didn't even come with a > hard drive, altho a "PC expansion unit" with a whopping 10M hard drive > could be had. Bzzzt. There was no expansion unit and no hard drive, until the PC-XT and PC-2 came out. ...yes I have an expansion unit for my "PC-1" also. It also has an 8087 and 704K of memory, though these are not "stock". ;-) > My book doesn't mention video (_Upgrading and Repairing PCs_), but I'm > pretty sure CGA was available as an option with the original. Correct. I have one: original (first day order) PC with both a monochrome *and* a CGA. I have both tubes too, though I bought the color tube a year or so later. Both were "available" from day-one though. ---- Keith ###### From: Bruce Hoult Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Tue, 22 Aug 2000 22:58:32 +1200 Organization: The Internet Group Ltd Lines: 17 Message-ID: References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> NNTP-Posting-Host: p57-max6.wlg.ihug.co.nz User-Agent: MT-NewsWatcher/3.0 (PPC) 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!ihug.co.nz!bruce Xref: chonsp.franklin.ch alt.folklore.computers:62362 In article , Eric Smith wrote: > When discussing "32-bit color", usually one means 8 bits per RGB > component. Maynard means that the eye can perceive an intensity > difference of 1 in 2**8. > > Others have claimed that they eye cannot distinguish 2*24 colors, so > 8 bits per component is good enough. Maynard points out (correctly, > IMNSHO) that it is not. A lot of this depends on whether you're talking a linear or a log scale. On the usual linear scale, the various shades of "nearly black" are quite easily distinguishable on an 8-bit ramp. And this is where games such as doom/quake/marathon spend most of their time. -- Bruce ###### From: Sander Vesik Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 22 Aug 2000 11:09:30 GMT Lines: 35 Message-ID: <966942570.224230@haldjas.folklore.ee> References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> NNTP-Posting-Host: haldjas.folklore.ee Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 8bit X-Trace: kadri.ut.ee 966942570 19218 193.40.6.121 (22 Aug 2000 11:09:30 GMT) X-Complaints-To: usenet@news.ut.ee NNTP-Posting-Date: 22 Aug 2000 11:09:30 GMT User-Agent: tin/pre-1.4-19990413 ("Endemoniada") (UNIX) (FreeBSD/3.4-RC (i386)) Cache-Post-Path: haldjas.folklore.ee!unknown@localhost.folklore.ee X-Cache: nntpcache 2.3.2 (see http://www.nntpcache.org/) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news.algonet.se!newsfeed1.telenordia.se!algonet!newsfeed1.funet.fi!newsfeeds.funet.fi!newsfeed.uninet.ee!news.ut.ee!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62334 In comp.arch Eric Smith wrote: > Maynard Handley wrote: >> It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp >> in 32 bits. > Del Cecchi writes: >> I don't do graphics, so please bear with me. Do you mean the eye can >> perceive an intensity difference of 1 in 2**32? > When discussing "32-bit color", usually one means 8 bits per RGB > component. Maynard means that the eye can perceive an intensity > difference of 1 in 2**8. > Others have claimed that they eye cannot distinguish 2*24 colors, so > 8 bits per component is good enough. Maynard points out (correctly, > IMNSHO) that it is not. Sa I have understood it, human eye isn't in any way linear - so that while in lot's of areas of grey the resolution of intensity levels is more than enough, in other areas you want about 2-4x more levels. The problem is made only worser if they are arranged as a ramp. And if you want to use an easy and naive encoding like 0-(2^n-1) there is nothing to do but increase number of levels everywhere. The problem is (imho) much more worse for green, speaking nothing of if you want green-blue blends... I'm sure there are people who deal with medical imaging in this group who can elaborate quite a bit more (and why 10 and 12 bit grey scale gets used). -- Sander FLW: "I can banish that demon" ###### Message-ID: <39A2608D.193D0A65@mikron.de> Date: Tue, 22 Aug 2000 13:14:21 +0200 From: Bernd Paysan X-Mailer: Mozilla 4.72 [de] (X11; I; Linux 2.2.14 i686) X-Accept-Language: de, en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 20 Organization: Customer of UUNET Deutschland GmbH NNTP-Posting-Host: 194.139.17.38 X-Trace: businessnews.de.uu.net 966945603 15175 194.139.17.38 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!blackbush.xlink.net!newsfeed01.sul.t-online.de!t-online.de!do.de.uu.net!nr-do2.de.uu.net!businessnews.de.uu.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62374 Eric Smith wrote: > When discussing "32-bit color", usually one means 8 bits per RGB > component. Maynard means that the eye can perceive an intensity > difference of 1 in 2**8. > > Others have claimed that they eye cannot distinguish 2*24 colors, so > 8 bits per component is good enough. Maynard points out (correctly, > IMNSHO) that it is not. The eye can't distinguish 2^24 colors, but it can distinguish 2^10 levels of gray. A 10:5:5 LCC should be completely sufficient, that's just 2^20 colors+intensities. That's what's representable on a CRT screen with limited contrast, if you want unlimited contrast, add 5 or 6 bits exponent to the luminance (that's then between "hurts" and "black cat in dark cave"). -- Bernd Paysan "If you want it done right, you have to do it yourself" http://www.jwdt.com/~paysan/ ###### From: Sander Vesik Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 22 Aug 2000 12:05:46 GMT Lines: 42 Message-ID: <966945945.99663@haldjas.folklore.ee> References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> NNTP-Posting-Host: haldjas.folklore.ee Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 8bit X-Trace: kadri.ut.ee 966945946 8710 193.40.6.121 (22 Aug 2000 12:05:46 GMT) X-Complaints-To: usenet@news.ut.ee NNTP-Posting-Date: 22 Aug 2000 12:05:46 GMT User-Agent: tin/pre-1.4-19990413 ("Endemoniada") (UNIX) (FreeBSD/3.4-RC (i386)) Cache-Post-Path: haldjas.folklore.ee!unknown@localhost.folklore.ee X-Cache: nntpcache 2.3.2 (see http://www.nntpcache.org/) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.icl.net!newspeer.clara.net!news.clara.net!news.algonet.se!newsfeed1.telenordia.se!algonet!newsfeed1.funet.fi!newsfeeds.funet.fi!newsfeed.uninet.ee!news.ut.ee!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62332 In comp.arch Bruce Hoult wrote: > In article , Eric Smith > wrote: >> When discussing "32-bit color", usually one means 8 bits per RGB >> component. Maynard means that the eye can perceive an intensity >> difference of 1 in 2**8. >> >> Others have claimed that they eye cannot distinguish 2*24 colors, so >> 8 bits per component is good enough. Maynard points out (correctly, >> IMNSHO) that it is not. > A lot of this depends on whether you're talking a linear or a log scale. > On the usual linear scale, the various shades of "nearly black" are > quite easily distinguishable on an 8-bit ramp. And this is where games > such as doom/quake/marathon spend most of their time. In some other world, everybody is probably using Lab or some even more advanced encoding and it all get's converted inside the RAMDAC which can negociate stuff like alpha conversion with the monitor by itself and always get it right. In this world, most things expect the encoding to be a number of bits of RGB and it's unclear what it would take to move to something better. But probably a lot as all software makes lot's of assumption about it, even if it isn't using strange tricks. Imagine all software using inline x86 asm (and by all, I do mean all - or rather more than 99.99%) and you wanted to make a new architecture to start competing with it - except all software has to be rewriteen as nobody ever even fancied another arch getting used outside labs. As regards to portability to other color models, most of software is really horrific, as nobody even thought of such things when writing it, making it totally horrific for porting. > -- Bruce -- Sander FLW: "I can banish that demon" ###### From: cecchi@signa.rchland.ibm.com (Del Cecchi) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 22 Aug 2000 13:01:32 GMT Organization: IBM Corp. Lines: 30 Message-ID: <8nttjc$116k$1@news.rchland.ibm.com> References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> Reply-To: dcecchi@vnet.ibm.com NNTP-Posting-Host: signa.rchland.ibm.com X-Trace: news.rchland.ibm.com 966949292 34004 9.5.54.183 (22 Aug 2000 13:01:32 GMT) X-Complaints-To: usenet@news.rchland.ibm.com NNTP-Posting-Date: 22 Aug 2000 13:01:32 GMT X-Newsreader: xrn 9.01-beta-3 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!bignews.mediaways.net!abq.news.ans.net!news.chips.ibm.com!newsfeed.btv.ibm.com!news!signa.rchland.ibm.com!cecchi Xref: chonsp.franklin.ch alt.folklore.computers:62346 In article , Eric Smith writes: |> Maynard Handley wrote: |> > It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp |> > in 32 bits. |> |> Del Cecchi writes: |> > I don't do graphics, so please bear with me. Do you mean the eye can |> > perceive an intensity difference of 1 in 2**32? |> |> When discussing "32-bit color", usually one means 8 bits per RGB |> component. Maynard means that the eye can perceive an intensity |> difference of 1 in 2**8. |> |> Others have claimed that they eye cannot distinguish 2*24 colors, so |> 8 bits per component is good enough. Maynard points out (correctly, |> IMNSHO) that it is not. |> |> I guess the "gray ramp in 32 bits" threw me. So it is really 8 bits for rgb and intensity? 256 gray levels of intensity might well be perceptable, especially given that the eye seems to be non linear, and moving from intensity=3 to intensity=4 is a 33 percent change. Or are display adapters suitably non linear also? -- Del Cecchi cecchi@rchland ###### Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <39A1DC28.A694DCBA@ibm.net> <966942570.224230@haldjas.folklore.ee> Organization: Massachvsetts Institvte of Technology From: jfc@mit.edu (John F Carr) Date: 22 Aug 2000 13:46:18 GMT Lines: 50 Message-ID: <39a2842a$0$9425@senator-bedfellow.mit.edu> NNTP-Posting-Host: NERD-XING.MIT.EDU X-Trace: dreaderd 966951978 9425 18.184.0.47 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.cwix.com!bloom-beacon.mit.edu!senator-bedfellow.mit.edu!dreaderd!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62358 In article <966942570.224230@haldjas.folklore.ee>, Sander Vesik wrote: > >Sa I have understood it, human eye isn't in any way linear - so that >while in lot's of areas of grey the resolution of intensity levels is >more than enough, in other areas you want about 2-4x more levels. The >problem is made only worser if they are arranged as a ramp. And if >you want to use an easy and naive encoding like 0-(2^n-1) there is >nothing to do but increase number of levels everywhere. > >I'm sure there are people who deal with medical imaging in this group >who can elaborate quite a bit more (and why 10 and 12 bit grey scale >gets used). I used to work in medical imaging. I did software*, not image science. Our product was greyscale. The eye has inferior color resolution, both in space and intensity. First, keep in mind that radiologists have better eyes than humans and they are trained to look for small intensity variations. Some of them can distinguish 1000 different intensities. The normal human eye is an 8-9 bit device. In either case the minimum perceived intensity change is nearly constant in logarithmic intensity. My eyes aren't good so I usually just accepted the statement that a particular sheet of film looked good or bad. Second, device response doesn't match eye response. Suppose your eyes can distinguish 2^10 density levels (density = -log_10 intensity) and you are looking for a disease that is visible around density 1.8. You want steps of .003 or less density at that particular point. Maybe it takes 12 bits to ensure that the critical area has adequate resolution. Our printer was write-white and was limited at high density by the smallest spot size we could draw and the smallest variation in spot size we could command. Suppose you can draw white areas from 10 square microns to 10000 square microns in 1 square micron steps within 100 micron square pixels. That gives you over 2^13 different intensity levels but it doesn't give you a linear 13 bit perceptual response. As I recall, conventional film has poor resolution at low density and good resolution at high density. *I debugged a programmable logic device program too. I don't know whether to call that software or hardware. The boundary used to be clear. -- John Carr (jfc@mit.edu) ###### From: mq-usenet@NOSPAMmaq.net (Michael Quinlan) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Reply-To: mq-usenet@NOSPAMmaq.org Message-ID: <39a387a8.4955818@news.micron.net> References: <39A1DC28.A694DCBA@ibm.net> <966942570.224230@haldjas.folklore.ee> <39a2842a$0$9425@senator-bedfellow.mit.edu> X-Newsreader: Forte Agent 1.5/32.452 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 10 Date: Tue, 22 Aug 2000 14:02:08 GMT NNTP-Posting-Host: 209.19.150.87 X-Trace: news03.micron.net 966952894 209.19.150.87 (Tue, 22 Aug 2000 08:01:34 MDT) NNTP-Posting-Date: Tue, 22 Aug 2000 08:01:34 MDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.cwix.com!skin02.micron.net!news03.micron.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62308 jfc@mit.edu (John F Carr) wrote: >First, keep in mind that radiologists have better eyes than humans... I always suspected that radiologists weren't human... -- No electrons were injured in the preparation of this message. Michael Quinlan mq@maq.org http://www.maq.org ###### Message-ID: <39A29BC2.AEA4C021@mikron.de> Date: Tue, 22 Aug 2000 17:26:58 +0200 From: Bernd Paysan X-Mailer: Mozilla 4.72 [de] (X11; I; Linux 2.2.14 i686) X-Accept-Language: de, en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> <8nttjc$116k$1@news.rchland.ibm.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 21 Organization: Customer of UUNET Deutschland GmbH NNTP-Posting-Host: 194.139.17.38 X-Trace: businessnews.de.uu.net 966960008 1878 194.139.17.38 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!uni-erlangen.de!newsfeed.germany.net!newsfeed01.sul.t-online.de!t-online.de!do.de.uu.net!nr-do2.de.uu.net!businessnews.de.uu.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62271 Del Cecchi wrote: > I guess the "gray ramp in 32 bits" threw me. So it is really 8 bits for rgb and > intensity? 256 gray levels of intensity might well be perceptable, especially > given that the eye seems to be non linear, and moving from intensity=3 to > intensity=4 is a 33 percent change. Or are display adapters suitably non linear > also? The display adapters are nonlinear, too (gamma=2.2). With a gray ramp (256 gray shades), you typically can see the Mach bands of all shades except the very light and very black ones (the black ones are hidden, because the black is really just dark gray, and the light ones are clipped). Mach bands don't say that you could distinguish these two colors if seen with a gap between, but that the contrast is large enough for the eye to be seen (the local contrasts are filtered directly in the eye, the brain only sees d/dt differences, and since the focus moves slowly over the point you are fixating, you get dx/dt or dy/dt). -- Bernd Paysan "If you want it done right, you have to do it yourself" http://www.jwdt.com/~paysan/ ###### From: michael.wojcik@merant.com (Michael Wojcik) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 22 Aug 2000 15:59:23 GMT Organization: MERANT Inc. Lines: 21 Message-ID: <8nu80r09df@news2.newsguy.com> References: <39A1DC28.A694DCBA@ibm.net> <966942570.224230@haldjas.folklore.ee> <39a2842a$0$9425@senator-bedfellow.mit.edu> <39a387a8.4955818@news.micron.net> Reply-To: michael.wojcik@merant.com NNTP-Posting-Host: p-118.newsdawg.com X-Newsreader: xrn 9.00 Originator: mww@lorelei.michaelwojcik.org 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!newsfeed.stanford.edu!pln-w!spln!dex!extra.newsguy.com!newsp.newsguy.com!mww Xref: chonsp.franklin.ch alt.folklore.computers:62315 In article <39a387a8.4955818@news.micron.net>, mq-usenet@NOSPAMmaq.net (Michael Quinlan) writes: > jfc@mit.edu (John F Carr) wrote: > >First, keep in mind that radiologists have better eyes than humans... > I always suspected that radiologists weren't human... I believe he meant that the eyes possessed by radiologists are better than the humans possessed by radiologists. This stands to reason, since radiologists generally get their humans from the hospital, where the stock tends to be sickly. -- Michael Wojcik michael.wojcik@merant.com AAI Development, MERANT (block capitals are a company mandate) Department of English, Miami University When most of what you do is a bit of a fraud, the word "profession" starts to look like the Berlin Wall. -- Tawada Yoko (t. M. Mitsutani) ###### From: jmfbahciv@aol.com Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 23 Aug 00 08:12:46 GMT Organization: UltraNet Communications, Inc. Lines: 54 Message-ID: <8o0boj$67m$4@bob.news.rcn.net> References: <8noeok$q6t$6@bob.news.rcn.net> <8nttge$eis$11@bob.news.rcn.net> <8nuvn2$299c$1@nntp1.ba.best.com> X-Trace: EbAsTrnG4rhm/wU2FXBKH+oee70ZL/Lxxn8mC8pP6AU= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 23 Aug 2000 11:15:31 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 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!feed1.news.rcn.net!rcn!207-172-255-186 Xref: chonsp.franklin.ch alt.folklore.computers:62406 alt.sys.pdp8:105 alt.sys.pdp10:194 alt.sys.pdp11:68 In article <8nuvn2$299c$1@nntp1.ba.best.com>, inwap@best.com (Joe Smith) wrote: >In article <8nttge$eis$11@bob.news.rcn.net>, wrote: >>In article , >> Eric Smith wrote: >>>jmfbahciv@aol.com writes: >>>> Now I'm not sure what you mean by architecture. Don't you >>>> consider the KA and the KL, both PDP-10s, different hardware >>>> architecture? We sure did. >>> >>>In the Blaauw & Brooks taxonomy [*], architecture is what is seen by the >>>programmers. In that sense, the PDP-6 and KA10 are nearly identical. >>>The later PDP-10 processors are mostly the same, with the addition >>>of paging, which is mostly only an issue for the monitor developers. >> >>Not at all. There were more differences than paging. One of >>design goals of each version of the processor was backwards >>compatibility. I don't think that necessarily confines a new architecture. >.... >>I wasn't considering the electronic components as a litmus test. > >The litmus test I use is the opcode table. >If 90% of the opcodes on the new machine are identical to the old machine, >then it is the same architecture. Pre-BIS, a compiler could treat a KL >as KA with a few extra double-precision instructions. Then a CPU whose hardware archtecture is foo-based but runs microcode that simulates a PDP-10 is which architecture? The actual hardware or the instruction set being executed? > >> I can't think of a single programming level >>that wasn't affected by each CPU implementation. I suppose >>it was mostly invisible at the compiler level in the cases >>of FORTRAN and COBOL. But I can think of aspects that even >>that level of programming changed, especially with advent >>of extended addressing on the KL. > >Yep, those were two things that through a monkey wrench into the simple >criteria I had been using. Adding the Business Instruction Set (EXTEND >with BCD arithemetic) to the KL made the system a different beast to the >COBOL compiler. And going from a single-section KL to extended addressing >had major ramifications to all compilers that wanted to use more than >256K-words. And I was thinking of all the "adjustments" that were required between CORE UUO and PAGE UUO when addressing mechanisms changed at the hardware level. /BAH Subtract a hundred and four for e-mail. ###### From: handleym@ricochet.net (Maynard Handley) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Tue, 22 Aug 2000 11:46:18 -0700 Organization: None Lines: 43 Message-ID: References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> NNTP-Posting-Host: handma2.apple.com X-Trace: news.apple.com 966969961 15259 17.202.32.152 (22 Aug 2000 18:46:01 GMT) X-Complaints-To: usenet@news.apple.com NNTP-Posting-Date: 22 Aug 2000 18:46:01 GMT X-Newsreader: MT-NewsWatcher 2.3.1 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!blackbush.xlink.net!howland.erols.net!newsfeed.direct.ca!look.ca!logbridge.uoregon.edu!newsfeed.stanford.edu!paloalto-snf1.gtei.net!news.gtei.net!forum.apple.com!news.apple.com!handma2.apple.com!user Xref: chonsp.franklin.ch alt.folklore.computers:62431 In article <39A1DC28.A694DCBA@ibm.net>, Del Cecchi wrote: >Maynard Handley wrote: >> >> In article <399D9B1D.1E0CE346@dusknet.dhs.org>, Jeff Teunissen >> wrote: >> >> >Nick Maclaren wrote: >> >> >> >> In article <39985C56.27AF@hda.hydro.com>, >> >> Terje Mathisen wrote: >> >[snip] >> >> >John Carmack wrote a long (.plan?) message this spring, calling for >> >> >64-bit graphics, using 16 bits (in fp format) for each of R, G & B, plus >> >> >another 16 bits for Alpha. >> >> >> >> Hmm. What species does he belong to? Mere humans cannot distinguish >> >> more than about 2^20 colour/intensity combinations .... >> > >> >It's an OpenGL thing. When you do things like applying shading, glow maps, >> >and so on to a base texture (particularly with variable alpha) in "only" >> >32bpp, banding is a common result. Doing it in 64 bits and sampling down to >> >32 (which is what Carmack's .plan was talking about) helps eliminate the >> >problem. >> >> It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp >> in 32 bits. >> >> Maynard > >I don't do graphics, so please bear with me. Do you mean the eye can >perceive an intensity difference of 1 in 2**32? Or do you mean >something else by grey ramp? Also it seems as if one would have only >about 1 tenmillionth of the ramp displayed because the screen is only on >the order of 1000 pixels out of 10 billion intensity levels. What I mean is that if you display a grey ramp on a 32-bit display you will see banding. Sure, the grey ramp (because it is grey) only utilizes 2^8 distinct "colors" but that is precisely the point---that a 32-bit display only allows for 8-bits per component, which means this sort of posterization occurs whenever one has color ramps. Maynard ###### From: Michael Lee Finney Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Message-ID: References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> <39A2608D.193D0A65@mikron.de> Reply-To: michael.finney@acm.org X-Newsreader: MicroPlanet Gravity v2.30 Lines: 35 Date: Tue, 22 Aug 2000 18:46:51 GMT NNTP-Posting-Host: 216.207.180.174 X-Complaints-To: Abuse Role , We Care X-Trace: monger.newsread.com 966970011 216.207.180.174 (Tue, 22 Aug 2000 14:46:51 EDT) NNTP-Posting-Date: Tue, 22 Aug 2000 14:46:51 EDT Organization: Lynchburg.net (lynchburg.net) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!xfer13.netnews.com!netnews.com!news-xfer.newsread.com!bad-news.newsread.com!netaxs.com!newsread.com!POSTED.monger.newsread.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62439 In article <39A2608D.193D0A65@mikron.de>, bpaysan@mikron.de says... > Eric Smith wrote: > > When discussing "32-bit color", usually one means 8 bits per RGB > > component. Maynard means that the eye can perceive an intensity > > difference of 1 in 2**8. > > > > Others have claimed that they eye cannot distinguish 2*24 colors, so > > 8 bits per component is good enough. Maynard points out (correctly, > > IMNSHO) that it is not. > > The eye can't distinguish 2^24 colors, but it can distinguish 2^10 > levels of gray. A 10:5:5 LCC should be completely sufficient, that's > just 2^20 colors+intensities. That's what's representable on a CRT > screen with limited contrast, if you want unlimited contrast, add 5 or 6 > bits exponent to the luminance (that's then between "hurts" and "black > cat in dark cave"). I remember hearing about a color blind person that had trained himself to distinguish something like 250,000 shades of gray. So, it may be possible to learn to distinguish 2**24 different colors. After all, that is only around 8 bits per individual color receptor which isn't too much to expect. Consider that if you were presented only with shades of pure red, could you learn to distinguish 350 different shades? Or 350 of green or 120 of blue? You have about half as many blue receptors as red or green because when red or green is focused (they are very close together) blue is fuzzy which is why "blue blocker" sunglasses work. I used to design my window colors scheme so that I used a blue background and all foreground colors were red, green or yellow. I could use significantly smaller text than with more conventional color design -- but it was so garish that people kept changing the screen colors! My point is don't confuse what the eye can perceive and what most people have learned to recognize. ###### From: ab528@FreeNet.Carleton.CA (Heinz W. Wiggeshoff) Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: 22 Aug 2000 19:04:33 GMT Organization: The National Capital FreeNet Lines: 12 Message-ID: <8nuis1$a5n$1@freenet9.carleton.ca> References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> <39A2608D.193D0A65@mikron.de> Reply-To: ab528@FreeNet.Carleton.CA (Heinz W. Wiggeshoff) NNTP-Posting-Host: freenet10 X-Trace: freenet9.carleton.ca 966971073 10423 134.117.136.30 (22 Aug 2000 19:04:33 GMT) X-Complaints-To: complaints@ncf.ca NNTP-Posting-Date: 22 Aug 2000 19:04:33 GMT X-Given-Sender: ab528@freenet10.carleton.ca (Heinz W. Wiggeshoff) 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!logbridge.uoregon.edu!newsfeed.stanford.edu!news.kjsl.com!xcski.com!freenet-news!FreeNet.Carleton.CA!ab528 Xref: chonsp.franklin.ch alt.folklore.computers:62423 Michael Lee Finney (michael.finney@acm.org) writes: > ... > Consider that if you were presented only with shades of pure red, could you > learn to distinguish 350 different shades? Or 350 of green or 120 of blue? ... Here in Canada, there's a hilarious commercial on TV in which a young couple, (OK, _she_) is trying to find a suitable colour of white for a room. "White is a hard colour..." Thank goodness my computer cabinetry is beige, or HP grey! ###### From: handleym@ricochet.net (Maynard Handley) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Tue, 22 Aug 2000 12:08:07 -0700 Organization: None Lines: 50 Message-ID: References: <39A1DC28.A694DCBA@ibm.net> <966942570.224230@haldjas.folklore.ee> <39a2842a$0$9425@senator-bedfellow.mit.edu> NNTP-Posting-Host: handma2.apple.com X-Trace: news.apple.com 966971270 15670 17.202.32.152 (22 Aug 2000 19:07:50 GMT) X-Complaints-To: usenet@news.apple.com NNTP-Posting-Date: 22 Aug 2000 19:07:50 GMT X-Newsreader: MT-NewsWatcher 2.3.1 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!newsfeed.stanford.edu!paloalto-snf1.gtei.net!news.gtei.net!forum.apple.com!news.apple.com!handma2.apple.com!user Xref: chonsp.franklin.ch alt.folklore.computers:62430 In article <39a2842a$0$9425@senator-bedfellow.mit.edu>, jfc@mit.edu (John F Carr) wrote: >In article <966942570.224230@haldjas.folklore.ee>, >Sander Vesik wrote: >> >>Sa I have understood it, human eye isn't in any way linear - so that >>while in lot's of areas of grey the resolution of intensity levels is >>more than enough, in other areas you want about 2-4x more levels. The >>problem is made only worser if they are arranged as a ramp. And if >>you want to use an easy and naive encoding like 0-(2^n-1) there is >>nothing to do but increase number of levels everywhere. >> >>I'm sure there are people who deal with medical imaging in this group >>who can elaborate quite a bit more (and why 10 and 12 bit grey scale >>gets used). > >I used to work in medical imaging. I did software*, not image science. >Our product was greyscale. The eye has inferior color resolution, both >in space and intensity. > >First, keep in mind that radiologists have better eyes than humans and >they are trained to look for small intensity variations. Some of them >can distinguish 1000 different intensities. The normal human eye is >an 8-9 bit device. In either case the minimum perceived intensity >change is nearly constant in logarithmic intensity. My eyes aren't good >so I usually just accepted the statement that a particular sheet of film >looked good or bad. This takes us off on a tangent, but to what extent does this actually matter when it comes to the real object of the exercise, namely diagnosis? How much of the extra content that is in an image created at some very high dpi and at 12 bpp is actually useful to diagnosis rather than simply noise? Or to put it another way, have studies been done that degraded the image in various ways (using some sense---eg don't just reduce to 8 bit, but reduce to 8 bit then fake the equiv of 12 bit through blue noise dither), and has it been shown that diagnosis was worse as a result? While I have utmost respect for the medical RESEARCH establishment, I have rather less respect for practicing medical types, and the claims made above strike me as of a kind with what one hears from them all the time---either untrue, or true but irrelevant. (Don't get me wrong---I'm not trying to push for some sort of political agenda here, and I am willing to accept that higher quality images may help in other ways, eg allowing for faster, albeit no more accurate, diagnosis. I'm simply curious these details.) Maynard ###### From: handleym@ricochet.net (Maynard Handley) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Tue, 22 Aug 2000 12:28:45 -0700 Organization: None Lines: 74 Message-ID: References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> <8nttjc$116k$1@news.rchland.ibm.com> NNTP-Posting-Host: handma2.apple.com X-Trace: news.apple.com 966972512 15832 17.202.32.152 (22 Aug 2000 19:28:32 GMT) X-Complaints-To: usenet@news.apple.com NNTP-Posting-Date: 22 Aug 2000 19:28:32 GMT X-Newsreader: MT-NewsWatcher 2.3.1 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!logbridge.uoregon.edu!newsfeed.stanford.edu!paloalto-snf1.gtei.net!news.gtei.net!forum.apple.com!news.apple.com!handma2.apple.com!user Xref: chonsp.franklin.ch alt.folklore.computers:62432 In article <8nttjc$116k$1@news.rchland.ibm.com>, dcecchi@vnet.ibm.com wrote: >In article , > Eric Smith writes: >|> Maynard Handley wrote: >|> > It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp >|> > in 32 bits. >|> >|> Del Cecchi writes: >|> > I don't do graphics, so please bear with me. Do you mean the eye can >|> > perceive an intensity difference of 1 in 2**32? >|> >|> When discussing "32-bit color", usually one means 8 bits per RGB >|> component. Maynard means that the eye can perceive an intensity >|> difference of 1 in 2**8. >|> >|> Others have claimed that they eye cannot distinguish 2*24 colors, so >|> 8 bits per component is good enough. Maynard points out (correctly, >|> IMNSHO) that it is not. >|> >|> > >I guess the "gray ramp in 32 bits" threw me. So it is really 8 bits for rgb and >intensity? 256 gray levels of intensity might well be perceptable, especially >given that the eye seems to be non linear, and moving from intensity=3 to >intensity=4 is a 33 percent change. Or are display adapters suitably non linear >also? Aah, you've opened up a real can of worms there with that question. This takes you into the whole area of "gamma" where one can flail around for days reading without encountering any sense. What you would like to happen is that the R, G, B values equal something that physical increases linearly (eg number of photons per sec, or radiant energy per sec). In fact what happens is that the display combo (monitor plus DACs) send the RGB values through a curve like energyOut=(R|G|B)^gamma(=2.2 usually). Thus the nominal values you are dealing with are actually(R|G|B)=energy^(-gamma). This is supposedly not only acceptable but great because it maps the available RGB values (0, 1, 2, ..255) to the energies in a way that better matches human perception. The extent to which is causes problems, in that addition does not correspond to adding energies, depends on what the addition is supposed to be achieving. Overall, given that everyone does this, it seems to mesh OK with what people see and it's only an issue if you are trying to mimic real physics in some way in your code. Even there, one can use internal RGB values in whatever format one likes (8-bit int if one wants, or floats), just as long as one does the right thing when it comes time to generate the RGB values for the video card. Where one has to be very careful is if one is trying to do something that is nominally physics based (I don' know, maybe a lens flare effect applied to a piece of video) that matches both physics and "normal" RGB data. However this is not the whole story. 2.2 has been the standard gamma forever because it matched the physics of early TV sets and associated video cameras. Apple, when they introduced color to the Mac in 1987 or so chose to have the monitor/DAC combo behave like a gamma of 1.8, because then the mapping of available RGB values to energies matches somewhat better the human eye. But, we have a whole world of images (JPEGs, GIFs) and movies (MPEGs, QT) which are generally not tagged with a gamma value and when displaying them you have no idea whether they were created to look good on a Mac (and need to be gamma corrected on PCs) or vice versa. MPEGs are a little better in that, in theory, they are targetted at 2.2 gamma and usually they are actually authored that way. QT movies can hold info describing the gamma for which they were authored (and QT will apply gamma correction as part of the blit as necessary, even when blitting through hardware YUV), but for that to work the authoring app must actually store the info describing that gamma, and must get it correct, both of which are somewhat iffy. When it comes to JPEG and GIF, you're basically doomed. Maynard ###### Sender: eric@ruckus.brouhaha.com From: Eric Smith Newsgroups: comp.arch,alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8nf2i9$2hgk$1@nntp1.ba.best.com> <399BA11E.BA1A907B@epl.ericsson.se> <40kupssgofs79lv592b74nsk8scrtvo1j5@4ax.com> X-Disclaimer: Everything I write is false. Organization: Eric Conspiracy Secret Labs X-Eric-Conspiracy: There is no conspiracy. Date: 22 Aug 2000 14:54:13 -0700 Message-ID: Lines: 9 X-Newsreader: Gnus v5.5/Emacs 20.3 NNTP-Posting-Host: ruckus.brouhaha.com X-Trace: 22 Aug 2000 15:00:29 -0700, ruckus.brouhaha.com Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!sunqbc.risq.qc.ca!news-hog.berkeley.edu!ucberkeley!enews.sgi.com!news.sgi.com!news.spies.com!ruckus.brouhaha.com Xref: chonsp.franklin.ch alt.folklore.computers:62426 On Mon, 21 Aug 2000 03:06:31, fixxit@bright.net wrote: > and 1 or 2 double density 5 1/4" drives (720k 3 1/2" was optional). krw@attglobal.net (Keith R. Williams) writes: > ..again wrong. The first PC had one or two double-density, but > *single*-sided 40-track, 8-sector/track (that means 140K per > diskette) drives. Bzzt! The first PC had *zero*, one, or two diskette drives. ###### From: mXrYmZp@email.com (Miha Peternel) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 23 Aug 2000 00:32:00 +0200 Organization: University of Maribor Lines: 24 Message-ID: References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> NNTP-Posting-Host: tp3.fov.uni-mb.si X-Trace: strelovod.uni-mb.si 966985255 26725 193.2.122.67 (22 Aug 2000 23:00:55 GMT) X-Complaints-To: news@uni-mb.si NNTP-Posting-Date: 22 Aug 2000 23:00:55 GMT X-Newsreader: MicroPlanet Gravity v2.10.940 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!npeer.kpnqwest.net!nmaster.kpnqwest.net!blackbush.xlink.net!newsfeed01.sul.t-online.de!t-online.de!newscore.univie.ac.at!kanja.arnes.si!news.uni-mb.si!locker Xref: chonsp.franklin.ch alt.folklore.computers:62425 In article , handleym@ricochet.net says... > >It's an OpenGL thing. When you do things like applying shading, glow maps, > >and so on to a base texture (particularly with variable alpha) in "only" > >32bpp, banding is a common result. Doing it in 64 bits and sampling down to > >32 (which is what Carmack's .plan was talking about) helps eliminate the > >problem. > > It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp > in 32 bits. The human eyes and brain are very good at distinguishing small differences in luminance. When there are two bands of very similar color/luminance, the eye will still notice the border between the bands. This is why we will need more than 8 bits per color component eventually. Of course, having even more bits helps when you want to do arithmetics on the components and not lose precision. Some people say 10 bits per component is enough, while others say at least 12 bits per component are needed. That should be enough for display, but for processing 16 bits come handy. Miha ###### From: inwap@best.com (Joe Smith) Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp8,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? Date: 22 Aug 2000 22:43:46 GMT Organization: Chez Inwap Lines: 40 Message-ID: <8nuvn2$299c$1@nntp1.ba.best.com> References: <8noeok$q6t$6@bob.news.rcn.net> <8nttge$eis$11@bob.news.rcn.net> NNTP-Posting-Host: shell3.ba.best.com X-Trace: nntp1.ba.best.com 966984226 75052 206.184.139.134 (22 Aug 2000 22:43:46 GMT) X-Complaints-To: abuse@best.com NNTP-Posting-Date: 22 Aug 2000 22:43:46 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!news.uni-stuttgart.de!logbridge.uoregon.edu!feeder.via.net!news1.best.com!nntp1.ba.best.com!inwap Xref: chonsp.franklin.ch alt.folklore.computers:62412 alt.sys.pdp8:107 alt.sys.pdp10:200 alt.sys.pdp11:70 In article <8nttge$eis$11@bob.news.rcn.net>, wrote: >In article , > Eric Smith wrote: >>jmfbahciv@aol.com writes: >>> Now I'm not sure what you mean by architecture. Don't you >>> consider the KA and the KL, both PDP-10s, different hardware >>> architecture? We sure did. >> >>In the Blaauw & Brooks taxonomy [*], architecture is what is seen by the >>programmers. In that sense, the PDP-6 and KA10 are nearly identical. >>The later PDP-10 processors are mostly the same, with the addition >>of paging, which is mostly only an issue for the monitor developers. > >Not at all. There were more differences than paging. One of >design goals of each version of the processor was backwards >compatibility. I don't think that necessarily confines a new architecture. ... >I wasn't considering the electronic components as a litmus test. The litmus test I use is the opcode table. If 90% of the opcodes on the new machine are identical to the old machine, then it is the same architecture. Pre-BIS, a compiler could treat a KL as KA with a few extra double-precision instructions. > I can't think of a single programming level >that wasn't affected by each CPU implementation. I suppose >it was mostly invisible at the compiler level in the cases >of FORTRAN and COBOL. But I can think of aspects that even >that level of programming changed, especially with advent >of extended addressing on the KL. Yep, those were two things that through a monkey wrench into the simple criteria I had been using. Adding the Business Instruction Set (EXTEND with BCD arithemetic) to the KL made the system a different beast to the COBOL compiler. And going from a single-section KL to extended addressing had major ramifications to all compilers that wanted to use more than 256K-words. -Joe -- See http://www.inwap.com/ for PDP-10 and "ReBoot" pages. ###### From: Dan.Pop@cern.ch (Dan Pop) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: 22 Aug 2000 23:39:28 GMT Organization: CERN European Lab for Particle Physics Lines: 24 Message-ID: <8nv2vg$eau$1@sunnews.cern.ch> References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8nf2i9$2hgk$1@nntp1.ba.best.com> <399BA11E.BA1A907B@epl.ericsson.se> <40kupssgofs79lv592b74nsk8scrtvo1j5@4ax.com> NNTP-Posting-Host: ues5.cern.ch X-Trace: sunnews.cern.ch 966987568 14686 (None) 137.138.32.79 X-Complaints-To: news@sunnews.cern.ch X-Newsreader: NN version 6.5.0 #18 (NOV) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!skynet.be!newsfeed.esat.net!EU.net!cern.ch!news Xref: chonsp.franklin.ch alt.folklore.computers:62410 In Eric Smith writes: >On Mon, 21 Aug 2000 03:06:31, fixxit@bright.net wrote: >> and 1 or 2 double density 5 1/4" drives (720k 3 1/2" was optional). > >krw@attglobal.net (Keith R. Williams) writes: >> ..again wrong. The first PC had one or two double-density, but >> *single*-sided 40-track, 8-sector/track (that means 140K per >> diskette) drives. > >Bzzt! The first PC had *zero*, one, or two diskette drives. And their capacity was 160k not 140k. 140k was the capacity of the 35-track drives (512 x 8 x 40 = 160k, 512 x 8 x 35 = 140k). A later DOS version (1.1?) introduced the 9-sector/track format, which lead to 180k for a single sided drive and 360k for a double sided one. Dan -- Dan Pop CERN, IT Division Email: Dan.Pop@cern.ch Mail: CERN - IT, Bat. 31 1-014, CH-1211 Geneve 23, Switzerland ###### From: krw@attglobal.net (Keith R. Williams) Reply-To: krw@attglobal.net Message-ID: Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <%0Tl5.132491$lU5.900249@news1.rdc1.nj.home.com> <8nf2i9$2hgk$1@nntp1.ba.best.com> <399BA11E.BA1A907B@epl.ericsson.se> <40kupssgofs79lv592b74nsk8scrtvo1j5@4ax.com> <8nv2vg$eau$1@sunnews.cern.ch> X-Newsreader: ProNews/2 Version 1.50 ia093a MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8bit NNTP-Posting-Host: 32.100.110.137 Date: 23 Aug 2000 02:33:36 GMT X-Trace: 23 Aug 2000 02:33:36 GMT, 32.100.110.137 Organization: Global Network Services - Remote Access Mail & News Services Lines: 30 X-Complaints-To: abuse@prserv.net Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!sunqbc.risq.qc.ca!arclight.uoregon.edu!logbridge.uoregon.edu!newsfeed.us.ibm.net!ibm.net!news3.prserv.net!32.100.110.137 Xref: chonsp.franklin.ch alt.folklore.computers:62436 On Tue, 22 Aug 2000 23:39:28, Dan.Pop@cern.ch (Dan Pop) wrote: > In Eric Smith writes: > > >On Mon, 21 Aug 2000 03:06:31, fixxit@bright.net wrote: > >> and 1 or 2 double density 5 1/4" drives (720k 3 1/2" was optional). > > > >krw@attglobal.net (Keith R. Williams) writes: > >> ..again wrong. The first PC had one or two double-density, but > >> *single*-sided 40-track, 8-sector/track (that means 140K per > >> diskette) drives. > > > >Bzzt! The first PC had *zero*, one, or two diskette drives. Blush. Correct. It had a cassette interface. > And their capacity was 160k not 140k. 140k was the capacity of the > 35-track drives (512 x 8 x 40 = 160k, 512 x 8 x 35 = 140k). > > A later DOS version (1.1?) introduced the 9-sector/track format, which > lead to 180k for a single sided drive and 360k for a double sided one. Yikes, I can't multiply/divide. You are indeed correct. ...I wonder if the ol' girl still works. I'm kinda afraid to try because my wife will banish it from the house if it has no value. ;-) ---- Keith ###### Sender: kma@eris.bgo.nera.no Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> From: Ketil Z Malde Mail-Copies-To: never Message-ID: Lines: 17 User-Agent: Gnus/5.0803 (Gnus v5.8.3) XEmacs/20.4 (Emerald) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Wed, 23 Aug 2000 06:22:00 GMT NNTP-Posting-Host: 195.204.181.130 X-Complaints-To: abuse@enitel.no X-Trace: news.telia.no 967011720 195.204.181.130 (Wed, 23 Aug 2000 08:22:00 CEST) NNTP-Posting-Date: Wed, 23 Aug 2000 08:22:00 CEST Organization: Enitel Internet Public Access 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!uninett.no!newsfeed1.enitel.no!news.telia.no!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62434 Bruce Hoult writes: > On the usual linear scale, the various shades of "nearly black" are > quite easily distinguishable on an 8-bit ramp. And this is where games > such as doom/quake/marathon spend most of their time. Of course, the game could simply adjust gamma up somewhat for really dark scenes, since that's how the eye normally behaves (and real world lightning - i.e. your desk lamp - prevents it from happening when playing games on a computer screen) (Guess nVidia might as well fire all those engineers, since they're not needed anymore :-) -kzm -- If I haven't seen further, it is by standing in the footprints of giants ###### Sender: kma@eris.bgo.nera.no Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <39A1DC28.A694DCBA@ibm.net> <966942570.224230@haldjas.folklore.ee> <39a2842a$0$9425@senator-bedfellow.mit.edu> From: Ketil Z Malde Mail-Copies-To: never Message-ID: Lines: 21 User-Agent: Gnus/5.0803 (Gnus v5.8.3) XEmacs/20.4 (Emerald) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Wed, 23 Aug 2000 06:26:16 GMT NNTP-Posting-Host: 195.204.181.130 X-Complaints-To: abuse@enitel.no X-Trace: news.telia.no 967011976 195.204.181.130 (Wed, 23 Aug 2000 08:26:16 CEST) NNTP-Posting-Date: Wed, 23 Aug 2000 08:26:16 CEST Organization: Enitel Internet Public Access 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!newsfeed.icl.net!colt.net!newspeer.clara.net!news.clara.net!news.algonet.se!algonet!nntp.se.dataphone.net!news.powertech.no!nntp.newmedia.no!newsfeed1.enitel.no!news.telia.no!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62433 jfc@mit.edu (John F Carr) writes: > Our product was greyscale. The eye has inferior color resolution, both > in space and intensity. > First, keep in mind that radiologists have better eyes than humans and > they are trained to look for small intensity variations. Some of them > can distinguish 1000 different intensities. The normal human eye is > an 8-9 bit device. So perhaps it'd make sense to work with some other color scheme, YUV or what have you, where one (high precision, say 16 bit) component is intensity, and two more (low precision, 8 bit each?) gives you color (and the third color is derivable from this). Might make the hardware more complex though, and would almost certainly break some APIs. -kzm -- If I haven't seen further, it is by standing in the footprints of giants ###### Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <39A39A87.4408CDF0@hda.hydro.com> Organization: Massachvsetts Institvte of Technology From: jfc@mit.edu (John F Carr) Date: 23 Aug 2000 12:00:14 GMT Lines: 16 Message-ID: <39a3bcce$0$9429@senator-bedfellow.mit.edu> NNTP-Posting-Host: NERD-XING.MIT.EDU X-Trace: dreaderd 967032014 9429 18.184.0.47 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed.mathworks.com!bloom-beacon.mit.edu!senator-bedfellow.mit.edu!dreaderd!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62437 In article , Christian Bau wrote: [How to get instructions into the I cache without executing them.] >Their solution: Put the address of the instruction into the lr register >(link register). Do an integer divide, for example 3/1 - that takes a few >cycles. Do a conditional branch to the link register if the result is zero >- but add a flag to the instruction to predict the branch as taken! That >will force the instruction to be loaded into cache, but only executed >speculatively, so there are no side effects. This is similar to SPARC V9. The instruction prefetch instruction is "branch never, predict taken". -- John Carr (jfc@mit.edu) ###### Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <399D9B1D.1E0CE346@dusknet.dhs.org> Organization: Massachvsetts Institvte of Technology From: jfc@mit.edu (John F Carr) Date: 23 Aug 2000 12:12:09 GMT Lines: 23 Message-ID: <39a3bf99$0$9445@senator-bedfellow.mit.edu> NNTP-Posting-Host: NERD-XING.MIT.EDU X-Trace: dreaderd 967032729 9445 18.184.0.47 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!cpk-news-hub1.bbnplanet.com!nycmny1-snh1.gtei.net!news.gtei.net!bloom-beacon.mit.edu!senator-bedfellow.mit.edu!dreaderd!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62435 In article , Miha Peternel wrote: >The human eyes and brain are very good at distinguishing small >differences in luminance. When there are two bands of very similar >color/luminance, the eye will still notice the border between the bands. The medical printer I worked on had trouble with a visible border between two bands of identical intensity, or indistinguishably different intensities, due to differences in the sub-pixel patterns used to create grey from black and white. The details of precisely how to avoid such artifacts were considered valuable trade secrets. For example, if you draw two checkerboard patterns next to each other offset by 1 square you will see a line along the boundary where the pattern changes. This sort of effect persists even when the squares are too small to resolve. This isn't really a function of bits per pixel, however, or powers of 2. I sense a slight topic drift... -- John Carr (jfc@mit.edu) ###### From: jmfbahciv@aol.com Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: Thu, 24 Aug 00 08:17:54 GMT Organization: UltraNet Communications, Inc. Lines: 22 Message-ID: <8o30ef$4u$3@bob.news.rcn.net> References: <8njlvt$ngg$1@citadel.in.taronga.com> <399C2FBC.225CB272@jetnet.ab.ca> <8njpp5$pok$1@citadel.in.taronga.com> <39A372B6.8ED08F06@hda.hydro.com> X-Trace: w4MHXgUjmOrpC4+FP1Yx+v+tOqFzNsvY4LBitl1A9mo= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 24 Aug 2000 11:20:47 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newscore.univie.ac.at!newsfeed.Austria.EU.net!nmaster.kpnqwest.net!npeer.kpnqwest.net!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!207-172-255-135 Xref: chonsp.franklin.ch alt.folklore.computers:62478 In article , swaim@nol.net wrote: >[Newsgroups trimmed] > >In alt.folklore.computers Terje Mathisen wrote: >> OTOH, it would still be nice if our general desktop OS's would also be >> capable of hard real-time scheduling, I'm just afraid that this would >> almost immediately be abused by some application vendor: By setting all >> my threads as hard realtime, I will get better benchmark results than >> all the others! Yeah! > > NT claims to have a "real time" priority. I believe that it has a number >of limitations, such as no disk access (including paging). > I decided, while I was in CompUSA to meet an NT. It didn't even survive 2 minutes of the den mother test and I was having great trouble messing with that mouse substitute. /BAH Subtract a hundred and four for e-mail. ###### From: swaim@nol.net Subject: Re: Q: Why not (2^n)-bit? Newsgroups: alt.folklore.computers References: <8njlvt$ngg$1@citadel.in.taronga.com> <399C2FBC.225CB272@jetnet.ab.ca> <8njpp5$pok$1@citadel.in.taronga.com> <39A372B6.8ED08F06@hda.hydro.com> User-Agent: tin/pre-1.4-19990413 ("Endemoniada") (UNIX) (FreeBSD/3.2-RELEASE (i386)) Lines: 15 Message-ID: X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly X-Complaints-To: support@usenetserver.com NNTP-Posting-Date: Wed, 23 Aug 2000 14:09:41 EDT Organization: WebUseNet Corp http://www.usenetserver.com - Home of the fastest NNTP servers on the Net. Date: Wed, 23 Aug 2000 18:09:41 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!news.uni-stuttgart.de!news.rhrz.uni-bonn.de!RRZ.Uni-Koeln.DE!news.netcologne.de!newsfeed01.sul.t-online.de!newsfeed00.sul.t-online.de!t-online.de!diablo.theplanet.net!cyclone2.usenetserver.com!news-out.usenetserver.com!cyclone1.usenetserver.com!news-west.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62452 [Newsgroups trimmed] In alt.folklore.computers Terje Mathisen wrote: > OTOH, it would still be nice if our general desktop OS's would also be > capable of hard real-time scheduling, I'm just afraid that this would > almost immediately be abused by some application vendor: By setting all > my threads as hard realtime, I will get better benchmark results than > all the others! Yeah! NT claims to have a "real time" priority. I believe that it has a number of limitations, such as no disk access (including paging). -- Mike Swaim, Avatar of Chaos: Disclaimer:I sometimes lie. Home: swaim at nol * net Quote: "Boingie"^4 Y,W&D ###### From: ddotpowell@netnospamscapeonline.co.uk (David Powell) Newsgroups: alt.folklore.computers,comp.arch,alt.sys.pdp10,alt.sys.pdp11,vmsnet.pdp-11 Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 23 Aug 2000 20:07:18 GMT Reply-To: ddotpowell@netscapeonline.co.uk Message-ID: <39a42ee5.5424532@newshost.netscapeonline.co.uk> References: <8n8mcb$ogt$1@yorikke.arb-phys.uni-dortmund.de> <39983350.2BF604F4@ev1.net> <39984D05.AB446489@elepar.com> <399851d6$1_1@news.wizvax.net> X-Newsreader: Forte Free Agent 1.21/32.243 NNTP-Posting-Host: userag43.netscapeonline.co.uk X-Trace: 23 Aug 2000 20:06:39 GMT, userag43.netscapeonline.co.uk Lines: 27 X-Report: Report abuse to abuse@netscapeonline.co.uk 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!newsfeed.icl.net!iclnet!plato.netscapeonline.co.uk!userag43.netscapeonline.co.uk Xref: chonsp.franklin.ch alt.folklore.computers:62447 alt.sys.pdp10:218 alt.sys.pdp11:73 On 14 Aug 2000 16:08:54 -0400, wilson@dbit.com (John Wilson) wrote: >In article <39984D05.AB446489@elepar.com>, >David C. DiNucci wrote: >Radix-50 (the 50 is octal of course, so it's 40.). > >>(In fact, I always figured it was this practice that led to >>conventions like 6-letter file names with 3-letter extensions, like >>those in early OSs.) > >Exactly. It also prevented the stupid punctuation in filenames that we >have to put up with now. > Sorry, it did prevent one stupidity, but not that one. "FO ..." "F O O " and " FO..O" are all valid RAD50 6 char strings. The punctuation had to be filtered out before the RAD50 pack. The stupidity it prevented was case sensitive filenames. And if we used RAD50 packing for the symbol table of assemblers and compilers, it stopped it there as well! Regards, David P. ###### From: "Thomas Womack" Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 23 Aug 2000 22:01:38 +0100 Organization: Customer of Energis Squared Lines: 25 Message-ID: <8o1l3j$5r5$3@newsg1.svr.pol.co.uk> References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> <39A2608D.193D0A65@mikron.de> NNTP-Posting-Host: modem-253.magnesium.dialup.pol.co.uk X-Trace: newsg1.svr.pol.co.uk 967071667 5989 62.136.11.253 (23 Aug 2000 23:01:07 GMT) NNTP-Posting-Date: 23 Aug 2000 23:01:07 GMT X-Complaints-To: abuse@theplanet.net X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2919.6600 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6600 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!nntp-out.monmouth.com!newspeer.monmouth.com!europa.netcrusader.net!208.184.7.66!newsfeed.skycache.com!Cidera!diablo.theplanet.net!news.theplanet.net!newspost.theplanet.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62497 "Bernd Paysan" wrote in message news:39A2608D.193D0A65@mikron.de... > Eric Smith wrote: > > When discussing "32-bit color", usually one means 8 bits per RGB > > component. Maynard means that the eye can perceive an intensity > > difference of 1 in 2**8. > > > > Others have claimed that they eye cannot distinguish 2*24 colors, so > > 8 bits per component is good enough. Maynard points out (correctly, > > IMNSHO) that it is not. > > The eye can't distinguish 2^24 colors, but it can distinguish 2^10 > levels of gray. A 10:5:5 LCC should be completely sufficient, that's > just 2^20 colors+intensities. Awfully unpleasant to do colour blends in that model, which is unfortunate since you're doing two or three of them per textured pixel, and many tens of millions of textured pixels per second. [Won't 10:5:5 look horribly banded on, say, a green-to-magenta colour blend? Though that's not especially pleasant in 24-bit either] Tom ###### From: Alexandre Pechtchanski Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Organization: Rockefeller University Hospital (GCRC), New York Message-ID: References: <8njlvt$ngg$1@citadel.in.taronga.com> <399C2FBC.225CB272@jetnet.ab.ca> <8njpp5$pok$1@citadel.in.taronga.com> <39A372B6.8ED08F06@hda.hydro.com> X-Newsreader: Forte Agent 1.7/32.534 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 20 Date: Thu, 24 Aug 2000 13:44:06 -0400 NNTP-Posting-Host: 129.85.24.56 X-Trace: rockyd.rockefeller.edu 967139275 129.85.24.56 (Thu, 24 Aug 2000 13:47:55 EDT) NNTP-Posting-Date: Thu, 24 Aug 2000 13:47:55 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.rhrz.uni-bonn.de!fu-berlin.de!news.maxwell.syr.edu!newsfeed.nyu.edu!rockyd.rockefeller.edu!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62575 On Wed, 23 Aug 2000 18:09:41 GMT, swaim@nol.net wrote: >[Newsgroups trimmed] > >In alt.folklore.computers Terje Mathisen wrote: >> OTOH, it would still be nice if our general desktop OS's would also be >> capable of hard real-time scheduling, I'm just afraid that this would >> almost immediately be abused by some application vendor: By setting all >> my threads as hard realtime, I will get better benchmark results than >> all the others! Yeah! > > NT claims to have a "real time" priority. I believe that it has a number >of limitations, such as no disk access (including paging). If we start list all of the things NT claims to have or to do... we'll have to rename this ng to something like "alt.folklore.lying-through-the-teeth" -- [ When replying, remove *'s from address ] Alexandre Pechtchanski, Systems Manager, RUH, NY ###### From: Brian Inglis Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Fri, 25 Aug 2000 20:07:04 -0600 Organization: Systematic Software Reply-To: Brian.dot.Inglis@SystematicSw.ab.ca Message-ID: <694cqs8qsnvkh2c1o2u5q3p4u5i2qo54fn@4ax.com> References: <39b24906.14742994@news.remarq.com> <399e6286.815215@news.remarq.com> <%%0n5.6508$6E.1554478@ptah.visi.com> <399D0EC6.D5887C74@netinsight.se> <39A117EE.97111822@mchpDOTsiemens.de> <39A120C2.1CF5606E@netinsight.se> <8o40if$dgk@gap.cco.caltech.edu> X-Newsreader: Forte Agent 1.8/32.548 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit NNTP-Posting-Host: 207.148.143.200 X-Original-NNTP-Posting-Host: 207.148.143.200 X-Trace: 25 Aug 2000 20:07:06 -0700, 207.148.143.200 Lines: 58 X-Original-NNTP-Posting-Host: 204.50.1.43 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!sunqbc.risq.qc.ca!nntp.cadvision.com!news.cadvision.com!207.148.143.200 Xref: chonsp.franklin.ch alt.folklore.computers:62607 On 24 Aug 2000 20:29:03 GMT, gah@ugcs.caltech.edu (glen herrmannsfeldt) wrote: >jmaynard@thebrain.conmicro.cx (Jay Maynard) writes: > >>On Mon, 21 Aug 2000 12:29:56 GMT, Johnny Billquist >> wrote: >>>No, it's an OS problem, since the OS design normally includes a software >>>timer. A hardware timer, even if it exists, is normally not used. > >>Counterexample: MVS and later IBM OSes. While the 360 didn't include a >>hardware time-of-day clock, and so OS/360 had to keep one based on the >>interval timer, the 370 added one as standard, and 370-based OSes use it >>exclusively. The interval timer was deleted in 370/XA. The hardware clock is >>only set at IPL time; changing the clock value at any time after that only >>adjusts an offset value that is added to the TOD clock return value before >>it's used by the OS. > >Counter counterexample. > >Linux/390 does not, as I understand, use the hardware timer. No -- it has to use the real TOD clock, and from what I've read, it seems to use a virtual clock comparator provided by VM, unless it can also run native on the hardware, in which case it must use the real clock comparator. There is only one timing mechanism on recent 370/390 -- 64 bit TOD clock and clock comparator. The TOD clock may only be set and started by the OS -- it then increments at a constant rate comparable to the instruction execution rate, but such that (big endian) bit 51 gets incremented every microsecond (bit 31 about every second). The limit of the architecture is then 1/4096us or about .25ns. After that, they need a new architecture, or need to relax the restrictions on the clock rate. The clock comparator can be set (only in privileged mode) to generate a clock comparator interrupt when the TOD clock reaches the value set in the comparator. Note also that the clock accuracy is only guaranteed to about one part in 100,000, so it can gain/lose about a second a day. And it's normally set by the OS at IPL to an operator entered value on a confirming keystroke, so most systems are off by seconds or even minutes, unless you fiddle with the OS software offset value, and always use OS calls to get the time. >And note that even with the S/360 interval timer, it would continue >to count if you were slow on the interrupt. It can count to over six >hours at 50 or 60 Hz, and interrupts when it goes through zero. >You can load a new value and retrieve the old value such that no counts >are lost. > >-- glen [pdp NGs trimmed] Thanks. Take care, Brian Inglis Calgary, Alberta, Canada -- Brian_Inglis@CSi.com (Brian dot Inglis at SystematicSw dot ab dot ca) use address above to reply ###### Message-ID: <39A875A3.B548A192@ibm.net> Date: Sat, 26 Aug 2000 20:57:55 -0500 From: Del Cecchi X-Mailer: Mozilla 4.7 [en]C-CCK-MCD NSCPCD47 (Win95; I) X-Accept-Language: en MIME-Version: 1.0 Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <8o3ogo$ji4$1@slb1.atl.mindspring.net> <39A7357A.8E713045@ev1.net> <8o7ju4$jvf$1@murrow.corp.sgi.com> <8o8idl$b6r$1@lisa.gemair.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit NNTP-Posting-Host: 32.100.32.197 X-Trace: 27 Aug 2000 01:58:26 GMT, 32.100.32.197 Organization: Global Network Services - Remote Access Mail & News Services Lines: 107 X-Complaints-To: abuse@prserv.net Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeeds.belnet.be!news.belnet.be!xfer13.netnews.com!netnews.com!cpk-news-hub1.bbnplanet.com!news.gtei.net!newsfeed.us.ibm.net!ibm.net!news3.prserv.net!32.100.32.197 Xref: chonsp.franklin.ch alt.folklore.computers:62635 Followups trimmed. we don't need every dec newsgroup in the universe crossposting. Jordan Henderson wrote: > > In article <8o7ju4$jvf$1@murrow.corp.sgi.com>, > John R. Mashey wrote: > >In article <39A7357A.8E713045@ev1.net>, Charles Richmond writes: > >|> Duane Sand wrote: > >|> > > >|> > [snip...] [snip...] [snip...] > >|> > > >|> > Yes, Bell's interview says that DEC's plan to eliminate all product lines > >|> > except for VAXes and low-end PDP-11 came from him, well after VAX > >|> > was introduced; it came to him during a Tahiti vacation. > >|> > > >|> Now all the 36-bitters from DEC know who to blame for the rape of the 36-bit > >|> line. Or maybe they already knew...I thought I detected some negativity from > >|> that group when Gordon Bell was discussed. Certainly I was disappointed in > >|> Bell when I heard this...seems he stabbed a lot of 36-bit customers in the > >|> back and lost a lot of paying business for DEC. > > > >1) This is hardly new news. > > > >2) The 36-bit fans were rightly devoted to their systems ... but, > >after all, this decisions was not unlike IBM's decision to stop doing > >36-bitters in favor of 32-bitters, and while comp.arch regularly beats > >up on the VAX ... it did make a *lot* of money for DEC, and it was the > >basis for what was probably DEC's best set of years, and it certainly > >caused trouble for DEC's competitors. > > > > Isn't this something of a false dichotomy? You make it sound as if > DEC would not have had the success with the VAX had they not abandoned > the 36-bit line. > > You should have a _very_ good reason to abandon a profitable line > like the DEC 36-bit line had been. I can tell you that few of the > 36-bit customers became loyal VAX customers. Some of them dabbled > in VAXen, but, from what I saw, these were some of the first to go > with other offerings during the VAXs declining years. A logical reason would be that it was not growing at a rapid enough rate to justify the required investment of precious development resource. > > The comparison with IBM is interesting. I don't know, but is it > true that the 36-bit line was abandoned by IBM for strategic > reasons alone? Or were these products actually becoming untenable? This is a tiny bit before my time, but it was strategic I have been told. I have been through several convergence crusades. Some have worked. The 360 one was before my time, but worked by the skin of IBM's teeth. > > I think you'll find that the DEC 36-bit line still had important > market niches when they were abandoned. So did the 7094 and 1401. > > Also, IBM has always had multiple overlapping product lines that > competed with each other. This decision by Gordon Bell was to > narrow the DEC product lines such that no product competed with > another. Bzzt. In the 360 era there was but one product line. Eventually some dumb farmers on the tundra sneaked the System/3 out by calling it a controller for punch card machines, and living in such a hostile environment that no executives would come to see. :-) 40 below does keep out the riff raff. > > >3) These are fairly strong & emotionally-loaded words from Charles ... > >now, the question for Charles is: to speak so strongly, I assume *you* have > >been in the position (like Gordon) of making > >difficult big decisions for your company, especially the > >always tough ones of having to rationalize a plethora of incompatible > >product lines, with each one having passionate champions, in the long-term > >hope of a better product line. Could you perhaps describe these > >decisions, in order to lend some credibility for your comments? > > Isn't this an ad-hominem? You know that Charles probably has > never held an executive position with a large computer > manufacturer, so you attack his credibility on this basis? Of course. Some (deleted perjorative adjectives and nouns) comes along 10 or 15 years after the fact whining about the decision to kill some machine he liked for some reason when he has never had to make those kind of decisions. Sort of like me complaining about the play selection of the Vikings on Monday Morning. Talk is cheap. > > This is similar to the arguments you hear from Veterans that you > have no credibility in criticizing military policy if you haven't > faught in wars. When actually, it's exactly the people who have been > in the heat of making hard decisions who can benefit best from > an outside perspective. Nah, it's loudmouths who have never experienced the reality who have the illusion that they could do better. > > >-- > >-John Mashey EMAIL: mash@sgi.com DDD: 650-933-3090 FAX: 650-933-2663 > >USPS: SGI 1600 Amphitheatre Pkwy., ms. 562, Mountain View, CA 94043-1351 > >SGI employee 25% time, non-conflicting,local, consulting elsewise. > > -Jordan Henderson > jordan@greenapple.com Del Cecchi ###### From: glass2@glass2.lexington.ibm.com Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 28 Aug 2000 13:12:59 GMT Organization: IBM Austin Lines: 44 Message-ID: <8odogr$fc2$1@ausnews.austin.ibm.com> References: <39b24906.14742994@news.remarq.com> <399e6286.815215@news.remarq.com> <%%0n5.6508$6E.1554478@ptah.visi.com> <399D0EC6.D5887C74@netinsight.se> <39A117EE.97111822@mchpDOTsiemens.de> <39A120C2.1CF5606E@netinsight.se> <8o40if$dgk@gap.cco.caltech.edu> <694cqs8qsnvkh2c1o2u5q3p4u5i2qo54fn@4ax.com> Reply-To: wa4qal@vnet.ibm.com NNTP-Posting-Host: glass2.cv.lexington.ibm.com X-Newsreader: IBM NewsReader/2 2.0 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeeds.belnet.be!news.belnet.be!newsgate.cistron.nl!newsfeed.uk.ibm.net!ibm.net!hursley.ibm.com!news2atm.raleigh.ibm.com!ausnews.austin.ibm.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62672 In <694cqs8qsnvkh2c1o2u5q3p4u5i2qo54fn@4ax.com>, Brian Inglis writes: > >No -- it has to use the real TOD clock, and from what I've read, >it seems to use a virtual clock comparator provided by VM, unless >it can also run native on the hardware, in which case it must use >the real clock comparator. >There is only one timing mechanism on recent 370/390 -- 64 bit >TOD clock and clock comparator. >The TOD clock may only be set and started by the OS -- it then >increments at a constant rate comparable to the instruction >execution rate, but such that (big endian) bit 51 gets >incremented every microsecond (bit 31 about every second). The >limit of the architecture is then 1/4096us or about .25ns. After >that, they need a new architecture, or need to relax the >restrictions on the clock rate. >The clock comparator can be set (only in privileged mode) to >generate a clock comparator interrupt when the TOD clock reaches >the value set in the comparator. >Note also that the clock accuracy is only guaranteed to about one >part in 100,000, so it can gain/lose about a second a day. >And it's normally set by the OS at IPL to an operator entered >value on a confirming keystroke, so most systems are off by >seconds or even minutes, unless you fiddle with the OS software >offset value, and always use OS calls to get the time. >[pdp NGs trimmed] >Thanks. Take care, Brian Inglis Calgary, Alberta, Canada >-- >Brian_Inglis@CSi.com (Brian dot Inglis at SystematicSw dot ab dot ca) > use address above to reply Too late, we already extended it. The TOD clock is now 104 bits long. Bits 64-103 are visible if the Extended-TOD-Clock Facility is installed. See Chapter 4 of the ESA/390 Principles of Operation, SA22-7201-06 (or later). Dave P.S. Standard Disclaimer: I work for them, but I don't speak for them. ###### From: hawk Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 28 Aug 2000 19:36:44 GMT Organization: Debian GNU/Linux site Lines: 28 Message-ID: References: , Ketil Z Malde wrote: >Bruce Hoult writes: >> On the usual linear scale, the various shades of "nearly black" are >> quite easily distinguishable on an 8-bit ramp. And this is where games >> such as doom/quake/marathon spend most of their time. >Of course, the game could simply adjust gamma up somewhat for really >dark scenes, since that's how the eye normally behaves (and real world >lightning - i.e. your desk lamp - prevents it from happening when >playing games on a computer screen) Now if only my cable company could get this to work. Digital cable over coax seems to be a bad idea. I'm giving up, I think--the digital signal is noticably worse than the analog, and we don't watch the extra channels much, anyway. ANd it's the blacks that are particularly bad; so bad that they draw attention to themselves. _Pale Rider_ *really* needs its darks . . . hawk -- Richard E. Hawkins dochawk@psu.edu [regardless of where the message says it comes from] These opinions will not be those of Penn State until they pay my retainer. ###### From: "Allen J. Baum" Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Mon, 28 Aug 2000 12:54:38 -0700 Organization: Compaq Computer Lines: 20 Message-ID: References: <39b24906.14742994@news.remarq.com> <399e6286.815215@news.remarq.com> <%%0n5.6508$6E.1554478@ptah.visi.com> <399D0EC6.D5887C74@netinsight.se> <39A117EE.97111822@mchpDOTsiemens.de> <39A120C2.1CF5606E@netinsight.se> <8o40if$dgk@gap.cco.caltech.edu> <694cqs8qsnvkh2c1o2u5q3p4u5i2qo54fn@4ax.com> NNTP-Posting-Host: althea.pa.dec.com User-Agent: MT-NewsWatcher/3.0 (PPC) 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!logbridge.uoregon.edu!newsfeed.stanford.edu!paloalto-snf1.gtei.net!news.gtei.net!newsgate.tandem.com!mailint03.im.hou.compaq.com!pa.dec.com!src.dec.com!allen.baum Xref: chonsp.franklin.ch alt.folklore.computers:62643 Brian.dot.Inglis@SystematicSw.ab.ca wrote: >The TOD clock may only be set and started by the OS -- it then >increments at a constant rate comparable to the instruction >execution rate, but such that (big endian) bit 51 gets >incremented every microsecond (bit 31 about every second). The >limit of the architecture is then 1/4096us or about .25ns. After >that, they need a new architecture, or need to relax the >restrictions on the clock rate. So what is the execution rate of a 4-way superscalar running at a 1Ghz clock? 1 ns or .25 ns? -- ********************************************** * Allen J. Baum tel. (650)853-6626 * * Compaq Computer Corp. fax (650)853-6513 * * 181 Lytton Ave. * * Palo Alto, CA 95306 abaum@pa.dec.com * ********************************************** ###### Sender: kma@eris.bgo.nera.no Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: From: Ketil Z Malde Mail-Copies-To: never Message-ID: Lines: 38 User-Agent: Gnus/5.0803 (Gnus v5.8.3) XEmacs/20.4 (Emerald) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Tue, 29 Aug 2000 06:04:43 GMT NNTP-Posting-Host: 195.204.181.130 X-Complaints-To: abuse@enitel.no X-Trace: news.telia.no 967529083 195.204.181.130 (Tue, 29 Aug 2000 08:04:43 CEST) NNTP-Posting-Date: Tue, 29 Aug 2000 08:04:43 CEST Organization: Enitel Internet Public Access Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!news.tele.dk!128.39.3.166!uninett.no!news.powertech.no!nntp.newmedia.no!newsfeed1.enitel.no!news.telia.no!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62717 [Fixed the references: field, you really should contemplate getting a real newsreader] hawk writes: > Now if only my cable company could get this to work. Digital cable over > coax seems to be a bad idea. I don't think the medium has anything to do with it - oh, well, I guess the bandwidth limitation is what makes the cable companies squeeze quality to get lower bit-rates. > ..in response to what I wrote: >> Of course, the game could simply adjust gamma up somewhat for really >> dark scenes, since that's how the eye normally behaves (and real world >> lightning - i.e. your desk lamp - prevents it from happening when >> playing games on a computer screen) When I wrote the above, I was rather curious why this wasn't implemented anywhere, since it doesn't sound like a very hard thing to do. Thinking about it, I guess it really is the same as using more bits in the pipe, and rendering to 256^3 colors with a gamma adjustment for display. But, if we somehow were able to calculate the "gamma factor" in advance, shouldn't we be able to move the gamma adjustment out in front of the calculation, and achieve the same result with only a 24 bit rendering process? One simple way of getting the pre-calculated gamma, is of course to use the previous frame(s). Gives you a neat "blinding" effect when running from darkness to brightness, and an adjustment period when running back again. -kzm -- If I haven't seen further, it is by standing in the footprints of giants ###### From: gah@ugcs.caltech.edu (glen herrmannsfeldt) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 29 Aug 2000 20:43:50 GMT Organization: California Institute of Technology, Pasadena Lines: 43 Message-ID: <8oh7a6$62g@gap.cco.caltech.edu> References: <39b24906.14742994@news.remarq.com> <399e6286.815215@news.remarq.com> <%%0n5.6508$6E.1554478@ptah.visi.com> <399D0EC6.D5887C74@netinsight.se> <39A117EE.97111822@mchpDOTsiemens.de> <39A120C2.1CF5606E@netinsight.se> <8o40if$dgk@gap.cco.caltech.edu> <694cqs8qsnvkh2c1o2u5q3p4u5i2qo54fn@4ax.com> 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.nextra.ch!news1.sunrise.ch!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!xfer13.netnews.com!netnews.com!logbridge.uoregon.edu!nntp-server.caltech.edu!gah Xref: chonsp.franklin.ch alt.folklore.computers:62702 Brian Inglis writes: >On 24 Aug 2000 20:29:03 GMT, gah@ugcs.caltech.edu (glen >herrmannsfeldt) wrote: (snip) >>Counter counterexample. >> >>Linux/390 does not, as I understand, use the hardware timer. >No -- it has to use the real TOD clock, and from what I've read, >it seems to use a virtual clock comparator provided by VM, unless >it can also run native on the hardware, in which case it must use >the real clock comparator. >There is only one timing mechanism on recent 370/390 -- 64 bit >TOD clock and clock comparator. No, there is also the CPU timer. See: http://www.s390.ibm.com:80/bookmgr-cgi/bookmgr.cmd/BOOKS/DZ9AR004/4%2e6%2e4 (Snip regarding the TOD clock) >And it's normally set by the OS at IPL to an operator entered >value on a confirming keystroke, so most systems are off by >seconds or even minutes, unless you fiddle with the OS software >offset value, and always use OS calls to get the time. Sorry for the confusion. There is a "CPU timer" and a "Clock Comparator". Both are 64 bits long. (Though not all bits are used.) The CPU timer decrements and interrupts when negative. (Note, not just on positive to negative transition.) The clock comparator compares some bits against the TOD clock, and interrupts when the clock comparator is less than the corresponding bits of the TOD clock. It seems like the CPU timer has better resolution than the clock comparator. The clock comparator does not compare all bits of the TOD clock. See: http://www.s390.ibm.com:80/bookmgr-cgi/bookmgr.cmd/BOOKS/DZ9AR004/4%2e6%2e1 for the TOD clock and a little later for the Clock Comparator. -- glen ###### From: gah@ugcs.caltech.edu (glen herrmannsfeldt) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 29 Aug 2000 21:02:27 GMT Organization: California Institute of Technology, Pasadena Lines: 41 Message-ID: <8oh8d3$6si@gap.cco.caltech.edu> References: <8o3ogo$ji4$1@slb1.atl.mindspring.net> <39A7357A.8E713045@ev1.net> <8o7ju4$jvf$1@murrow.corp.sgi.com> <8o8idl$b6r$1@lisa.gemair.com> <39A875A3.B548A192@ibm.net> 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.ifi.unizh.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!logbridge.uoregon.edu!nntp-server.caltech.edu!gah Xref: chonsp.franklin.ch alt.folklore.computers:62705 Del Cecchi writes: >Followups trimmed. we don't need every dec newsgroup in the universe >crossposting. (snip some more) >> >2) The 36-bit fans were rightly devoted to their systems ... but, >> >after all, this decisions was not unlike IBM's decision to stop doing >> >36-bitters in favor of 32-bitters, and while comp.arch regularly beats >> >up on the VAX ... it did make a *lot* of money for DEC, and it was the >> >basis for what was probably DEC's best set of years, and it certainly >> >caused trouble for DEC's competitors. (and more) >> >> The comparison with IBM is interesting. I don't know, but is it >> true that the 36-bit line was abandoned by IBM for strategic >> reasons alone? Or were these products actually becoming untenable? Before 360 there were scientific machines (good at numeric stuff) and commercial machines (good at character stuff and simple arithmetic). Each incompatible with others of the same type (scientific or commercial). S/360 was supposed to do both, as well as be upward/downward compatible within the line. I forget how long they expected the architecture to last, but they might be surprised to find that S/360 software will still run on today's machines, over 30 years later. Note that this was also the beginning of the idea of a computer architecture, as different from the implementation. Also, they provided emulators (most with microcode assist) for many of the older processors, to support those users during the transition. >> I think you'll find that the DEC 36-bit line still had important >> market niches when they were abandoned. >So did the 7094 and 1401. Both of those had emulators in S/360 models. I wonder if DEC could have had a PDP-10 emulator that ran on VAX hardware? -- glen ###### From: cecchi@signa.rchland.ibm.com (Del Cecchi) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 29 Aug 2000 21:03:30 GMT Organization: IBM Corp. Lines: 44 Message-ID: <8oh8f2$13o4$1@news.rchland.ibm.com> References: <8o7ju4$jvf$1@murrow.corp.sgi.com> <8o896a$sea$2@bob.news.rcn.net> <8oah5l$rrb$1@mach.thp.univie.ac.at> <8oe2m0$rd1$1@nntp1.ba.best.com> <8og6ju$d8i$3@bob.news.rcn.net> Reply-To: dcecchi@vnet.ibm.com NNTP-Posting-Host: signa.rchland.ibm.com X-Trace: news.rchland.ibm.com 967583010 36612 9.5.54.183 (29 Aug 2000 21:03:30 GMT) X-Complaints-To: usenet@news.rchland.ibm.com NNTP-Posting-Date: 29 Aug 2000 21:03:30 GMT X-Newsreader: xrn 9.01-beta-3 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!informatik.tu-muenchen.de!news.informatik.uni-muenchen.de!uni-erlangen.de!newsfeeds.belnet.be!news.belnet.be!xfer13.netnews.com!netnews.com!cpk-news-hub1.bbnplanet.com!news.gtei.net!portc03.blue.aol.com!newsjunkie.ans.net!news.chips.ibm.com!newsfeed.btv.ibm.com!news!signa.rchland.ibm.com!cecchi Xref: chonsp.franklin.ch alt.folklore.computers:62713 In article <8og6ju$d8i$3@bob.news.rcn.net>, you write: |> In article <8oe2m0$rd1$1@nntp1.ba.best.com>, |> inwap@best.com (Joe Smith) wrote: snip (correctly I hope) (and followups trimmed |> > |> >The proper comparison, |> > VAX throughput per system * number of VAX systems at one site |> >against |> > PDP-10 throughput per system * number of PDP-10 systems at one site |> >showed the VAX did not have the computing horsepower that customers |> needed. |> > |> >A commonly quoted metric was that it took two VAX-8600's and a 780 to |> >equal the same amount of processing power as a single KL. And at the time |> >of the abandonment, no number of VAXes could match the job load of a |> >quad-SMP KL system. (VAXclusters != SMP). |> |> Yeah! That's telling them, Joe. ;-) |> |> I view that nonsense (VAX/VMS could replace a TOPS system's |> service throughput) as small computer thinking. These were |> not people who knew what timesharing was. |> |> /BAH That's just what the guys in Poughkeepsie used to say about those "toy computers" with those funny Intel processors and no RAS features. They kept on saying it for ten years, until about 1992 as I recall. It is that kind of thinking that got IBM where it is today in the PC business, the networking business, and the enterprise storage business, among other segments. -- Del Cecchi These are my personal opinions only, as if you couldn't figure that out. -- Del Cecchi cecchi@rchland ###### From: "Don Chiasson" Newsgroups: alt.folklore.computers,alt.sys.pdp10,alt.sys.pdp11 Subject: Re: Q: Why not (2^n)-bit? Date: Tue, 29 Aug 2000 21:34:45 -0400 Organization: Posted via Supernews, http://www.supernews.com Lines: 49 Message-ID: References: <39A7357A.8E713045@ev1.net> <8o7ju4$jvf$1@murrow.corp.sgi.com> <8o896a$sea$2@bob.news.rcn.net> <8oah5l$rrb$1@mach.thp.univie.ac.at> X-Complaints-To: newsabuse@supernews.com X-Newsreader: Microsoft Outlook Express 4.71.1712.3 X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!newsfeed-zh.ip-plus.net!news.ip-plus.net!news.tesion.net!news.belwue.de!news-stu1.dfn.de!news-mue1.dfn.de!news-nue1.dfn.de!uni-erlangen.de!newsfeeds.belnet.be!news.belnet.be!newsfeed.icl.net!newspeer.clara.net!news.clara.net!news.inweb.net.uk!news-1.opaltelecom.net!sn-xit-02!supernews.com!sn-inject-01!corp.supernews.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62743 alt.sys.pdp10:325 alt.sys.pdp11:113 Jonathan Thornburg wrote in message <8oah5l$rrb$1@mach.thp.univie.ac.at>... >In article <8o7ju4$jvf$1@murrow.corp.sgi.com>, >mash@mash.engr.sgi.com (John R. Mashey) wrote: >>2) The 36-bit fans were rightly devoted to their systems ... but, >>after all, this decisions was not unlike IBM's decision to stop doing >>36-bitters in favor of 32-bitters, and while comp.arch regularly beats >>up on the VAX ... it did make a *lot* of money for DEC, and it was the >>basis for what was probably DEC's best set of years, and it certainly >>caused trouble for DEC's competitors. > >In article <8o896a$sea$2@bob.news.rcn.net>, wrote: >>I suggest that you do some more analyses. The VAXes running >>VMS never replaced the computer power of a TOPS-10 SMP system, >>where "computer power" is measured by users' work that gets >>accomplished per unit of time. > >Oh really? If you're using that unit of measure, then we're comparing > VAX throughput per system * number of VAX systems >against > PDP-10 throughput per system * number of PDP-10 systems >I think the number of VAX systems extant was a *lot* larger than the >number of PDP-10 systems. I was involved in buying a DEC-20 in late 1977, and it was serial number 256. Extrapolating based on this, I would guess that perhaps 1,000 were DEC-20's were manufactured?? Don ###### From: Toon Moene Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 30 Aug 2000 18:01:11 +0200 Organization: Moene Computational Physics, Maartensdijk, The Netherlands Lines: 13 Message-ID: <39AD2FC7.68A2F27C@moene.indiv.nluug.nl> References: <8o3ogo$ji4$1@slb1.atl.mindspring.net> <39A7357A.8E713045@ev1.net> <8o7ju4$jvf$1@murrow.corp.sgi.com> <8o8idl$b6r$1@lisa.gemair.com> <39A875A3.B548A192@ibm.net> <8oh8d3$6si@gap.cco.caltech.edu> NNTP-Posting-Host: mail.moene.indiv.nluug.nl Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: porthos.nl.uu.net 967657653 2335 195.109.255.217 (30 Aug 2000 17:47:33 GMT) X-Complaints-To: abuse@nl.uu.net NNTP-Posting-Date: 30 Aug 2000 17:47:33 GMT X-Mailer: Mozilla 4.07 [en] (X11; I; Linux 2.2.12 i686) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!psinet-eu-nl!newsfeeds.belnet.be!news.belnet.be!newsgate.cistron.nl!skynet.be!newsfeed2.news.nl.uu.net!sun4nl!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62744 glen herrmannsfeldt wrote: > I wonder if DEC could have had a PDP-10 emulator that ran on VAX > hardware? Perhaps it's useful to know that a PDP-10 backend is discussed on the GCC mailing lists as we speak. -- Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290 Saturnushof 14, 3738 XG Maartensdijk, The Netherlands GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html GNU Fortran 95: http://g95.sourceforge.net/ (under construction) ###### From: jeff_diamond@my-deja.com Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 30 Aug 2000 19:17:26 GMT Organization: Deja.com - Before you buy. Lines: 54 Message-ID: <8ojmjf$3kj$1@nnrp1.deja.com> References: <966272518.763493@haldjas.folklore.ee> <39985C56.27AF@hda.hydro.com> <8n9p9l$6no$1@pegasus.csx.cam.ac.uk> <399D9B1D.1E0CE346@dusknet.dhs.org> <39A1DC28.A694DCBA@ibm.net> <8nttjc$116k$1@news.rchland.ibm.com> NNTP-Posting-Host: 24.4.252.75 X-Article-Creation-Date: Wed Aug 30 19:17:26 2000 GMT X-Http-User-Agent: Mozilla/4.74 [en]C-AtHome0407 (Win95; U) X-Http-Proxy: HTTP/1.1 avnl1.nj.home.com[18038023] (Traffic-Server/3.0.7 [uScMs f p eN:t cCMi p s ]), 1.0 x71.deja.com:80 (Squid/1.1.22) for client 24.4.252.75 X-MyDeja-Info: XMYDJUIDjeff_diamond 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!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!nntp2.deja.com!nnrp1.deja.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62760 In article , handleym@ricochet.net (Maynard Handley) wrote: > In article <8nttjc$116k$1@news.rchland.ibm.com>, dcecchi@vnet.ibm.com wrote: > > >In article , > > Eric Smith writes: > >|> Maynard Handley wrote: > >|> > It's NOT just an openGL thing. Dammit, you can see banding on a grey ramp > >|> > in 32 bits. > >|> > >|> Del Cecchi writes: > >|> > I don't do graphics, so please bear with me. Do you mean the eye can > >|> > perceive an intensity difference of 1 in 2**32? > >|> > >|> When discussing "32-bit color", usually one means 8 bits per RGB > >|> component. Maynard means that the eye can perceive an intensity > >|> difference of 1 in 2**8. > >|> > >|> Others have claimed that they eye cannot distinguish 2*24 colors, so > >|> 8 bits per component is good enough. Maynard points out (correctly, > >|> IMNSHO) that it is not. > >|> I have done a lot of research in this area, and here's what I've found. The human eye can distinguish about 512 shades of "perfectly calibrated" logirthmic intensity steps. (More in green, less in red.) So 8-bit color channels are pretty close, and uif calibrated correctly, you'll only see banding in a dark room and a good monitor. But since good calibration is rather rare, to be safe, 10-bit color channels will look perfect to the human eye, and have the fun side benefit of being just about 1,000 per channel. The reason Macs and professional video software use 16-bit channels (48-bit color), and why TIFF and NeXTs use 12-bit channels isn't because the eye is that good. It's so they can do color and intensity processing without losing fidelity. This is similar to scanning an image at 2400 dpi so you can later enlarge part of it and still view it at 400 DPI. Hope this helps. - Jeff Sent via Deja.com http://www.deja.com/ Before you buy. ###### From: cecchi@signa.rchland.ibm.com (Del Cecchi) Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: 30 Aug 2000 21:26:40 GMT Organization: IBM Corp. Lines: 65 Message-ID: <8oju6g$126i$1@news.rchland.ibm.com> References: <8o7ju4$jvf$1@murrow.corp.sgi.com> <8o896a$sea$2@bob.news.rcn.net> <8oah5l$rrb$1@mach.thp.univie.ac.at> <8oe2m0$rd1$1@nntp1.ba.best.com> <8og6ju$d8i$3@bob.news.rcn.net> <8oh8f2$13o4$1@news.rchland.ibm.com> <8oit7l$9bs$3@bob.news.rcn.net> Reply-To: dcecchi@vnet.ibm.com NNTP-Posting-Host: signa.rchland.ibm.com X-Trace: news.rchland.ibm.com 967670800 35026 9.5.54.183 (30 Aug 2000 21:26:40 GMT) X-Complaints-To: usenet@news.rchland.ibm.com NNTP-Posting-Date: 30 Aug 2000 21:26:40 GMT X-Newsreader: xrn 9.01-beta-3 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!newsgate.cistron.nl!newsfeed.uk.ibm.net!ibm.net!hursley.ibm.com!news2atm.raleigh.ibm.com!newsfeed.btv.ibm.com!news!signa.rchland.ibm.com!cecchi Xref: chonsp.franklin.ch alt.folklore.computers:62777 In article <8oit7l$9bs$3@bob.news.rcn.net>, jmfbahciv@aol.com writes: |> In article <8oh8f2$13o4$1@news.rchland.ibm.com>, |> cecchi@signa.rchland.ibm.com (Del Cecchi) wrote: |> >In article <8og6ju$d8i$3@bob.news.rcn.net>, |> > you write: |> >|> In article <8oe2m0$rd1$1@nntp1.ba.best.com>, |> >|> inwap@best.com (Joe Smith) wrote: |> >snip (correctly I hope) (and followups trimmed |> >|> > |> >|> >The proper comparison, |> >|> > VAX throughput per system * number of VAX systems at one site |> >|> >against |> >|> > PDP-10 throughput per system * number of PDP-10 systems at one site |> >|> >showed the VAX did not have the computing horsepower that customers |> >|> needed. |> >|> > |> >|> >A commonly quoted metric was that it took two VAX-8600's and a 780 to |> >|> >equal the same amount of processing power as a single KL. And at the |> time |> >|> >of the abandonment, no number of VAXes could match the job load of a |> >|> >quad-SMP KL system. (VAXclusters != SMP). |> >|> |> >|> Yeah! That's telling them, Joe. ;-) |> >|> |> >|> I view that nonsense (VAX/VMS could replace a TOPS system's |> >|> service throughput) as small computer thinking. These were |> >|> not people who knew what timesharing was. |> >|> |> >|> /BAH |> > |> >That's just what the guys in Poughkeepsie used to say |> >about those "toy computers" |> >with those funny Intel processors and no RAS features. |> >They kept on saying it |> >for ten years, until about 1992 as I recall. |> > |> >It is that kind of thinking that got IBM where it |> >is today in the PC business, |> >the networking business, and the enterprise storage business, among other |> >segments. |> |> I suspect that you have misunderstood my comment. By small |> computer thinking, I mean that the developers make certain |> assumptions that on the order of "this will never happen", |> completely precluding any extension of functionality. The |> key is to try not to program oneself into a corner. |> |> /BAH |> |> Subtract a hundred and four for e-mail. I understood exactly what you said and meant. And I have heard it before in almost those same words from the mid hudson valley and the shores of the Bay in San Jose. Replace TOPS with MVS and replace VMS with Unix for example. Or replace TOPS with 3380 and VMS with commodity pc drives. PDP10 is dead fred and so is DEC. They made one too many strategic mistakes or had one too many cases of bad luck. If IBM can go from monopoly threat to walking dead in a few years, it can happen to anyone. Just do enough things wrong. It often has little to do with technical merit. -- Del Cecchi cecchi@rchland ###### Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? References: <8oh8f2$13o4$1@news.rchland.ibm.com> <8oit7l$9bs$3@bob.news.rcn.net> <8oju6g$126i$1@news.rchland.ibm.com> Organization: None X-Newsreader: trn 4.0-test72 (19 April 1999) From: kragen@dnaco.net (Kragen Sitaker) Lines: 29 Message-ID: <1G%r5.8697$f65.465858@news-west.usenetserver.com> X-Abuse-Info: Please be sure to forward a copy of ALL headers X-Abuse-Info: Otherwise we will be unable to process your complaint properly X-Complaints-To: support@usenetserver.com NNTP-Posting-Date: Sat, 02 Sep 2000 00:50:05 EDT Date: Sat, 02 Sep 2000 04:50:05 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.imp.ch!sunqbc.risq.qc.ca!cyclone2.usenetserver.com!news-out.usenetserver.com!cyclone1.usenetserver.com!news-west.usenetserver.com.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:63007 In article <8oju6g$126i$1@news.rchland.ibm.com>, Del Cecchi wrote: >PDP10 is dead fred and so is DEC. They made one too many strategic mistakes or >had one too many cases of bad luck. If IBM can go from monopoly threat to >walking dead in a few years, it can happen to anyone. Just do enough things >wrong. It often has little to do with technical merit. DEC was doing all right --- not riding high like they were in the 1970s and 1980s --- but then they tangled with Intel. Patents are a serious problem; many extremely broad patents are being granted, so we end up with a few huge companies like Intel, IBM, and AMD owning the whole high-performance CPU industry, simply because they have patents broad enough to keep out nearly anyone else, and they have cross-licensing deals with one another. Fuck with Intel --- as DEC and Intergraph did --- and they will sink your boat. This is a terrible perversion of the original purpose of the patent system: to foster innovation. Today, it protects big players; little guys like Transmeta can only play if they make deals with one of the giants. In the near future, we will see the same thing happening with software, at least in the USA. -- Kragen Sitaker Perilous to all of us are the devices of an art deeper than we ourselves possess. -- Gandalf the Grey [J.R.R. Tolkien, "Lord of the Rings"] ###### From: "Bill Todd" Newsgroups: alt.folklore.computers,comp.arch Subject: Re: Q: Why not (2^n)-bit? Date: Sun, 3 Sep 2000 15:47:16 -0400 Organization: Why won't Outlook let me leave this blank? Lines: 28 Message-ID: <8ou9lg$c25$1@pyrite.mv.net> References: <8oh8f2$13o4$1@news.rchland.ibm.com> <8oit7l$9bs$3@bob.news.rcn.net> <8oju6g$126i$1@news.rchland.ibm.com> <1G%r5.8697$f65.465858@news-west.usenetserver.com> NNTP-Posting-Host: bnh-2-44.mv.com X-Trace: pyrite.mv.net 968010224 12357 199.125.99.108 (3 Sep 2000 19:43:44 GMT) X-Complaints-To: abuse@mv.com NNTP-Posting-Date: 3 Sep 2000 19:43:44 GMT X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2314.1300 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2314.1300 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!newspeer1.nac.net!news.mv.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:62847 Kragen Sitaker wrote in message news:1G%r5.8697$f65.465858@news-west.usenetserver.com... > In article <8oju6g$126i$1@news.rchland.ibm.com>, > Del Cecchi wrote: > >PDP10 is dead fred and so is DEC. They made one too many strategic mistakes or > >had one too many cases of bad luck. If IBM can go from monopoly threat to > >walking dead in a few years, it can happen to anyone. Just do enough things > >wrong. It often has little to do with technical merit. > > DEC was doing all right --- not riding high like they were in the 1970s > and 1980s --- but then they tangled with Intel. And, IIRC, more or less won: unloaded at a premium price an expensive fab that DEC couldn't make very profitable, plus made Intel commit to building Alphas until DEC didn't care any more whether they did. DEC's problems were vision and management, dating from the mid-'80s if not earlier, not Intel (except for the indirect influence of the PC and DEC's failure to understand its implications and turn them to DEC's advantage). - bill ###### From: Arargh! Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: Mon, 04 Sep 2000 22:46:39 -0500 Organization: Arargh!! Lines: 17 Message-ID: References: <2000Aug14.184433.8359@lorelei.approve.se> <39984e80$1_1@news.wizvax.net> <2000Aug14.212932.9879@lorelei.approve.se> <399895cd$1_1@news.wizvax.net> NNTP-Posting-Host: 207-229-148-155.d.enteract.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: news.enteract.com 968125386 28805 207.229.148.155 (5 Sep 2000 03:43:06 GMT) X-Complaints-To: abuse@enteract.com NNTP-Posting-Date: 5 Sep 2000 03:43:06 GMT X-Newsreader: Forte Agent 1.8/32.548 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!enews.sgi.com!korova.insync.net!newsfeed.enteract.com!news.enteract.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:63449 On 14 Aug 2000 20:58:53 -0400, wilson@dbit.com (John Wilson) wrote: >In article <2000Aug14.212932.9879@lorelei.approve.se>, >Goran Larsson wrote: >Again you're thinking about modern machines. The PDP-8 (dunno about other >12-bit machines like the ND-812 or whatever else there was) didn't have any The ND-812 didn't. >instructions which took single-bit operands, except for the ones which >operated on the LINK bit (like a carry bit), and there's only one of those. >There aren't enough bits to have any of the modern-day fanciness of encoding >complicated operands into the opcode, and even if they were the 8 wouldn't >have used them, since it was designed to be a cheap Model T type computer. -- arargh (at enteract period com) http://www.arargh.com (Reply address points nowhere in an attempt to foil e-mail spammers.) ###### From: Arargh! Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 06 Sep 2000 03:35:31 -0500 Organization: Arargh!! Lines: 27 Message-ID: References: <2000Aug14.184433.8359@lorelei.approve.se> <39984e80$1_1@news.wizvax.net> <2000Aug14.212932.9879@lorelei.approve.se> <399895cd$1_1@news.wizvax.net> NNTP-Posting-Host: 207-229-148-161.d.enteract.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: news.enteract.com 968228986 45110 207.229.148.161 (6 Sep 2000 08:29:46 GMT) X-Complaints-To: abuse@enteract.com NNTP-Posting-Date: 6 Sep 2000 08:29:46 GMT X-Newsreader: Forte Agent 1.8/32.548 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!rz.uni-karlsruhe.de!schlund.de!newsfeed01.sul.t-online.de!newsfeed00.sul.t-online.de!t-online.de!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.enteract.com!news.enteract.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:63418 On Mon, 04 Sep 2000 22:46:39 -0500, Arargh! wrote: >On 14 Aug 2000 20:58:53 -0400, wilson@dbit.com (John Wilson) wrote: > >>In article <2000Aug14.212932.9879@lorelei.approve.se>, >>Goran Larsson wrote: > >>Again you're thinking about modern machines. The PDP-8 (dunno about other >>12-bit machines like the ND-812 or whatever else there was) didn't have any >The ND-812 didn't. >>instructions which took single-bit operands, except for the ones which >>operated on the LINK bit (like a carry bit), and there's only one of those. I just found my 'Nuclear Data, Inc' Instruction list for the ND-812. It seems the only single bit things are the flag & overflow bits. >>There aren't enough bits to have any of the modern-day fanciness of encoding >>complicated operands into the opcode, and even if they were the 8 wouldn't >>have used them, since it was designed to be a cheap Model T type computer. > -- arargh (at enteract period com) http://www.arargh.com (Reply address points nowhere in an attempt to foil e-mail spammers.) ###### From: Arargh! Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: Wed, 06 Sep 2000 20:19:16 -0500 Organization: Arargh!! Lines: 12 Message-ID: <17rdrscpiueaoodmol6r2rjknliaihmkgc@4ax.com> References: <39a013e5.25745326@news.remarq.com> <399941CC.1CC6DFA5@netinsight.se> <3999ECF7.FCCFD121@bellatlantic.net> NNTP-Posting-Host: 207-229-148-41.d.enteract.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: news.enteract.com 968289245 15240 207.229.148.41 (7 Sep 2000 01:14:05 GMT) X-Complaints-To: abuse@enteract.com NNTP-Posting-Date: 7 Sep 2000 01:14:05 GMT X-Newsreader: Forte Agent 1.8/32.548 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-mue1.dfn.de!news-nue1.dfn.de!uni-erlangen.de!newsfeeds.belnet.be!news.belnet.be!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.enteract.com!news.enteract.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:63045 On Wed, 16 Aug 2000 01:14:59 GMT, hg/jb wrote: >pdp 16 was also called CADET - Can't Add, Doesn't Even Try. >The 16 was intended, iirc, an industrial control device >taht could be used for machine control. I thought the IBM 1620 was the 'CADET' -- arargh (at enteract period com) http://www.arargh.com (Reply address points nowhere in an attempt to foil e-mail spammers.) ###### From: "Allen J. Baum" Newsgroups: alt.folklore.computers Subject: Re: Q: Why not (2^n)-bit? Date: Fri, 08 Sep 2000 15:32:33 -0700 Organization: Compaq Computer Lines: 55 Message-ID: References: <2000Aug14.184433.8359@lorelei.approve.se> <39984e80$1_1@news.wizvax.net> <2000Aug14.212932.9879@lorelei.approve.se> <399895cd$1_1@news.wizvax.net> NNTP-Posting-Host: althea.pa.dec.com User-Agent: MT-NewsWatcher/3.0 (PPC) 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!informatik.tu-muenchen.de!news.informatik.uni-muenchen.de!uni-erlangen.de!newsfeeds.belnet.be!news.belnet.be!xfer13.netnews.com!netnews.com!newsfeed.skycache.com!Cidera!cyclone2.usenetserver.com!news-out.usenetserver.com!easynews!uunet!dfw.uu.net!mailint03.im.hou.compaq.com!pa.dec.com!src.dec.com!allen.baum Xref: chonsp.franklin.ch alt.folklore.computers:63547 I've lost the original email, so I'm not sure what the question was, but here's an answer anyway: I think the TI-9xx series had a pile of instructions that worked on bitfields which could be any length. The bitfield instructions may have been primarily for doing IO (it was intended for process control, so these instructions were to things to individual bit IO.) The Tenet 210 (a 32 bit mini-mainframe of the early 70's - only two were sold) had a whole pile of bitfield instructions, including add and subtract. The IBM Stretch was a bit addressable machine, though I don't recall if it had bitfield instructions Of course, the PDP-10 had load / store bitfield ops In article , Arargh! wrote: >On Mon, 04 Sep 2000 22:46:39 -0500, Arargh! >wrote: > >>On 14 Aug 2000 20:58:53 -0400, wilson@dbit.com (John Wilson) wrote: >> >>>In article <2000Aug14.212932.9879@lorelei.approve.se>, >>>Goran Larsson wrote: >> >>>Again you're thinking about modern machines. The PDP-8 (dunno about other >>>12-bit machines like the ND-812 or whatever else there was) didn't have any >>The ND-812 didn't. > >>>instructions which took single-bit operands, except for the ones which >>>operated on the LINK bit (like a carry bit), and there's only one of those. > >I just found my 'Nuclear Data, Inc' Instruction list for the ND-812. >It seems the only single bit things are the flag & overflow bits. > > >>>There aren't enough bits to have any of the modern-day fanciness of encoding >>>complicated operands into the opcode, and even if they were the 8 wouldn't >>>have used them, since it was designed to be a cheap Model T type computer. >> > > -- >arargh (at enteract period com) http://www.arargh.com >(Reply address points nowhere in an attempt to foil e-mail spammers.) -- ********************************************** * Allen J. Baum tel. (650)853-6626 * * Compaq Computer Corp. fax (650)853-6513 * * 181 Lytton Ave. * * Palo Alto, CA 95306 abaum@pa.dec.com * **********************************************