From: phil@panix.com (Phil Gustafson) Newsgroups: alt.folklore.computers Subject: Lights Was: Tubes in computers Date: 9 Oct 2000 16:09:18 -0400 Organization: Once-famed Parquet Floor Lines: 27 Message-ID: <8rt8le$n0c$1@panix2.panix.com> References: <8rm095$a4l$1@nnrp1.deja.com> <4a0a4c51e5a__fake__address@127.0.0.1> <8rt6vh$kks$1@panix2.panix.com> <78F041D890F3D0D4.A91FBA2705AC865D.E3207CC8DBF6DAB7@lp.airnews.net> NNTP-Posting-Host: panix2.panix.com X-Trace: news.panix.com 971122159 20872 166.84.0.227 (9 Oct 2000 20:09:18 GMT) X-Complaints-To: abuse@panix.com NNTP-Posting-Date: 9 Oct 2000 20:09:18 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeeds.belnet.be!news.belnet.be!xfer13.netnews.com!netnews.com!howland.erols.net!panix!news.panix.com!panix2.panix.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:65683 jstewart@jkmicro.com writes: >Phil Gustafson wrote: >> >> The same thing from a different angle: Through the KA-10, DEC had a >> design rule requiring one indicator lamp per flip-flop. > >Are you sure? This may be true for the data path, but I suspect there >were many flip-flops in the timing and clock generation circuitry that >did not have lamps associated with them. > When the KA-10 was being designed, I was a technician responsible for laying out the light arrays on various peripheral controllers. I believe I assigned a light to each and every flip-flop. Of course that was 33 years ago, yadda yadda, but I'll stick to my statement till corrected. (Oh. No lights for the IC fast registers.) (Oh. The KA didn't have a clock per se.) I also got to lay out switch panels. I feel old. Phil -- )) (( Phil Gustafson Urban Legends FAQ: http://www.urbanlegends.com C|~~| Java FAQ: http://www.afu.com `--' ###### From: jmfbahciv@aol.com Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: Wed, 11 Oct 00 07:45:46 GMT Organization: UltraNet Communications, Inc. Lines: 44 Message-ID: <8s1h03$hif$5@bob.news.rcn.net> References: <8rm095$a4l$1@nnrp1.deja.com> <4a0a4c51e5a__fake__address@127.0.0.1> <8rt6vh$kks$1@panix2.panix.com> <78F041D890F3D0D4.A91FBA2705AC865D.E3207CC8DBF6DAB7@lp.airnews.net> <8rt8le$n0c$1@panix2.panix.com> <39E32A6E.D11BB5DA@prescienttech.com> X-Trace: G7/RUfIgUvUFl0pCtrw4zqHNgYf4UF5fUHj2Txouwrw= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 11 Oct 2000 10:56:03 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!207-172-216-173 Xref: chonsp.franklin.ch alt.folklore.computers:65925 In article <39E32A6E.D11BB5DA@prescienttech.com>, "Carl R. Friend" wrote: >Phil Gustafson wrote: >> >> jstewart@jkmicro.com writes: >> >Phil Gustafson wrote: >> >> >> >> The same thing from a different angle: Through the KA-10, DEC >> >> had a design rule requiring one indicator lamp per flip-flop. >> > >> >Are you sure? This may be true for the data path, but I suspect >> >there were many flip-flops in the timing and clock generation >> >circuitry that did not have lamps associated with them. > > This trend continued on with the KI-10, which had lamps for pretty >much each flop in the machine (save, as Phil mentions, the Fast Memory >registers - the 36 lamps looked at the output) including the time- >state flops. When the KL was introduced, the large indicator panels >of the earlier machines were gone. > > Yes, the KA lacked a clock (asynchronous machine) but the KI had >one (110 nS basic cycle, extendable) but didn't have any lights >associated with it. With the KI, though, you could single-step the >time-states. All right...now remember that hardware is guy stuff ;-). I don't understand what you mean by the KA not having a clock. I always thought it did especially since a customer, who wanted to do real time, had to add on a real time clock and build the monitor with the realtime module. > >> I feel old. > > I think most of us do.... > Nah, only when I bend over. /BAH Subtract a hundred and four for e-mail. ###### From: "Carl R. Friend" Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: Tue, 10 Oct 2000 10:40:46 -0400 Organization: as little as possible! Lines: 34 Message-ID: <39E32A6E.D11BB5DA@prescienttech.com> References: <8rm095$a4l$1@nnrp1.deja.com> <4a0a4c51e5a__fake__address@127.0.0.1> <8rt6vh$kks$1@panix2.panix.com> <78F041D890F3D0D4.A91FBA2705AC865D.E3207CC8DBF6DAB7@lp.airnews.net> <8rt8le$n0c$1@panix2.panix.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: gI5qY8l4PM0ZK/gwnv9m1DokbxbmyO0sXyz7smem7A4= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 10 Oct 2000 14:40:48 GMT X-Accept-Language: en X-Mailer: Mozilla 4.5 [en] (X11; I; Linux 2.0.29 i586) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:65804 Phil Gustafson wrote: > > jstewart@jkmicro.com writes: > >Phil Gustafson wrote: > >> > >> The same thing from a different angle: Through the KA-10, DEC > >> had a design rule requiring one indicator lamp per flip-flop. > > > >Are you sure? This may be true for the data path, but I suspect > >there were many flip-flops in the timing and clock generation > >circuitry that did not have lamps associated with them. This trend continued on with the KI-10, which had lamps for pretty much each flop in the machine (save, as Phil mentions, the Fast Memory registers - the 36 lamps looked at the output) including the time- state flops. When the KL was introduced, the large indicator panels of the earlier machines were gone. Yes, the KA lacked a clock (asynchronous machine) but the KI had one (110 nS basic cycle, extendable) but didn't have any lights associated with it. With the KI, though, you could single-step the time-states. > I feel old. I think most of us do.... -- +------------------------------------------------+---------------------+ | Carl Richard Friend (UNIX Sysadmin) | West Boylston | | Minicomputer Collector / Enthusiast | Massachusetts, USA | | mailto:crfriend@ma.ultranet.com +---------------------+ | http://www.ultranet.com/~crfriend/museum | ICBM: 42:22N 71:47W | +------------------------------------------------+---------------------+ ###### From: phil@panix.com (Phil Gustafson) Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: 11 Oct 2000 16:33:51 -0400 Organization: Once-famed Parquet Floor Lines: 32 Message-ID: <8s2irf$5fi$1@panix2.panix.com> References: <8rm095$a4l$1@nnrp1.deja.com> <8rt8le$n0c$1@panix2.panix.com> <39E32A6E.D11BB5DA@prescienttech.com> <8s1h03$hif$5@bob.news.rcn.net> NNTP-Posting-Host: panix2.panix.com X-Trace: news.panix.com 971296431 4012 166.84.0.227 (11 Oct 2000 20:33:51 GMT) X-Complaints-To: abuse@panix.com NNTP-Posting-Date: 11 Oct 2000 20:33:51 GMT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!machtgarnix.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsxfer.eecs.umich.edu!panix!news.panix.com!panix2.panix.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:65930 jmfbahciv@aol.com writes: >All right...now remember that hardware is guy stuff ;-). > >I don't understand what you mean by the KA not having a clock. >I always thought it did especially since a customer, who wanted >to do real time, had to add on a real time clock and build >the monitor with the realtime module. > Well, simplifying a bit for gurls ;-): Most computers these days are synchronous. There's a clock that counts cadence for the whole cpu. Every few nanoseconds, there comes a pulse that initiates a state change in the system. This state change is (usually) known to be complete before the next pulse comes along to initiate the next change. The KA-10 had no such clock. When some part of the system changed state, it sent a signal to the next part. The signal said, "Hey, I have some data for you on my output bus". The next part accepted the data and sent back an acknowledment: "Thanks for the data. You may go about your business". There were certain advantages to this approach, but obviously not enough. Phil -- )) (( Phil Gustafson Urban Legends FAQ: http://www.urbanlegends.com C|~~| Java FAQ: http://www.afu.com `--' ###### From: jmfbahciv@aol.com Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: Thu, 12 Oct 00 08:12:43 GMT Organization: UltraNet Communications, Inc. Lines: 49 Message-ID: <8s46v0$9pf$3@bob.news.rcn.net> References: <8rm095$a4l$1@nnrp1.deja.com> <8rt8le$n0c$1@panix2.panix.com> <39E32A6E.D11BB5DA@prescienttech.com> <8s1h03$hif$5@bob.news.rcn.net> <8s2irf$5fi$1@panix2.panix.com> X-Trace: 5fbZzx87Q7hxIJupXzA1b8Bsl8hfUnyK8O3w9AHX2P4= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 12 Oct 2000 11:23:12 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!xfer13.netnews.com!netnews.com!feed1.news.rcn.net!rcn!207-172-216-66 Xref: chonsp.franklin.ch alt.folklore.computers:65977 In article <8s2irf$5fi$1@panix2.panix.com>, phil@panix.com (Phil Gustafson) wrote: >jmfbahciv@aol.com writes: >>All right...now remember that hardware is guy stuff ;-). >> >>I don't understand what you mean by the KA not having a clock. >>I always thought it did especially since a customer, who wanted >>to do real time, had to add on a real time clock and build >>the monitor with the realtime module. >> >Well, simplifying a bit for gurls ;-): Thanks. I need all the help I can get in this area. > >Most computers these days are synchronous. There's a clock that counts >cadence for the whole cpu. Every few nanoseconds, there comes a pulse >that initiates a state change in the system. This state change is >(usually) known to be complete before the next pulse comes along to >initiate the next change. > >The KA-10 had no such clock. When some part of the system changed >state, it sent a signal to the next part. The signal said, "Hey, I have >some data for you on my output bus". The next part accepted the data >and sent back an acknowledment: "Thanks for the data. You may go about >your business". When you whisper these words in my ear, I immediately think about hardware interrupts. But that's not what you are talking about... or is it? Anyway, something had to keep a count of the ticks. Wouldn't that have been called the clock? [red-cheeked emoticon hangs its head, having just figured out it's reached a conclusion based on associating a noun with a behaviour erroneously]. In my non-guy head, an accumulation of tick counts meant that a clock was going "tick, tick, tick, ...., tick". You can tell that I got most of my hardware education by listening over the wall. > >There were certain advantages to this approach, but obviously not >enough. Would you explain one advantage for those of us who are still learning? /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: Fri, 13 Oct 00 08:23:49 GMT Organization: UltraNet Communications, Inc. Lines: 57 Message-ID: <8s6s00$jpf$3@bob.news.rcn.net> References: <8rm095$a4l$1@nnrp1.deja.com> <8s1h03$hif$5@bob.news.rcn.net> <8s2irf$5fi$1@panix2.panix.com> <8s46v0$9pf$3@bob.news.rcn.net> <8s4ofk$9ip$1@panix2.panix.com> <%btF5.42119$pu4.4417427@typhoon.ne.mediaone.net> X-Trace: CT4qef3uxAqWgmiV7tAjnsSgB+sEo8ah5MYcetWmHGg= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 13 Oct 2000 11:34:24 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!207-172-102-66 Xref: chonsp.franklin.ch alt.folklore.computers:65986 In article <%btF5.42119$pu4.4417427@typhoon.ne.mediaone.net>, Ric Werme wrote: >phil@panix.com (Phil Gustafson) writes: > >>jmfbahciv@aol.com writes: > >>>When you whisper these words in my ear, I immediately think about >>>hardware interrupts. But that's not what you are talking about... >>>or is it? Anyway, something had to keep a count of the ticks. >>>Wouldn't that have been called the clock? > >>Ah. Misunderstanding. The clock I'm talking about has nothing to do >>with what time it is. It's the signal that keeps the computer in step >>with itself. The real-time clock is an independent gadget that works >>much like a digital wristwatch, but for some reason keeps worse time. > >While there was a separate real-time clock option for the -10 and places >like C-MU implemented their own, the clock Barb is thinking of was the >line clock. In the US, 60 hz, elsewhere 50 hz. Quite accurate, due >to the way power distribution has to work! That's it!!!! What a difference a name makes :-) Thanks. > >The clock Phil is talking about is the clock the steps a computer each >"quantum" of the way. On the -10, this sort of clock would tick several >times during a shift instruction, once for each shift. On the KA-10, >there were several things call "monostables" or "one-shots" where you'd >goose them with a pulse going in and they'd respond with a long pulse >coming out. They timed a lot of things from the shift rate to the delay >given on the I/O bus between asking for data and reading the bus. Whew! There's more to that hardware game than I had imagined. No wonder it was "guy thing" ;-). Weren't those pulses based on the line clock? Or did a CPU designer add a wire long enough so that it "timed" the speed of an electron through it. That would get awfully expensive if you had to have one/instruction and the timing was different for each instruction. > >Not too many things had the handshaking Phil describe, IIRC. The memory >system was the key one and allowed mixing memory of different speeds >together. Oh!!!!! [ emoticon here noticing that a light bulb just turned on]. Is that what all of that meant. Didn't we allow mixes but strongly recommended "don't do that"? IIRC, there was some talk about how the ordering of the boxes was a Very Important Thing. Now, I sorta think I know how TW did his disk and magtape timing, but would JMF have had to know the timing of the memory cycles in order to do his memory management and CPU drivers? /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: Fri, 13 Oct 00 08:08:52 GMT Organization: UltraNet Communications, Inc. Lines: 98 Message-ID: <8s6r40$jpf$1@bob.news.rcn.net> References: <8rm095$a4l$1@nnrp1.deja.com> <8s1h03$hif$5@bob.news.rcn.net> <8s2irf$5fi$1@panix2.panix.com> <8s46v0$9pf$3@bob.news.rcn.net> <8s4ofk$9ip$1@panix2.panix.com> X-Trace: 1JjXPzY4fGaavO7N/z62khboa9HdgRDHWP354/XxZEM= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 13 Oct 2000 11:19:28 GMT X-Newsreader: News Xpress Version 1.0 Beta #4 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!feed1.news.rcn.net!rcn!207-172-102-66 Xref: chonsp.franklin.ch alt.folklore.computers:65987 In article <8s4ofk$9ip$1@panix2.panix.com>, phil@panix.com (Phil Gustafson) wrote: >jmfbahciv@aol.com writes: >>In article <8s2irf$5fi$1@panix2.panix.com>, >>> >>>Most computers these days are synchronous. There's a clock that counts >>>cadence for the whole cpu. Every few nanoseconds, there comes a pulse >>>that initiates a state change in the system. This state change is >>>(usually) known to be complete before the next pulse comes along to >>>initiate the next change. >>> >>>The KA-10 had no such clock. When some part of the system changed >>>state, it sent a signal to the next part. The signal said, "Hey, I have >>>some data for you on my output bus". The next part accepted the data >>>and sent back an acknowledment: "Thanks for the data. You may go about >>>your business". >> >>When you whisper these words in my ear, I immediately think about >>hardware interrupts. But that's not what you are talking about... >>or is it? Anyway, something had to keep a count of the ticks. >>Wouldn't that have been called the clock? [red-cheeked emoticon >>hangs its head, having just figured out it's reached a conclusion >>based on associating a noun with a behaviour erroneously]. In my >>non-guy head, an accumulation of tick counts meant that a clock >>was going "tick, tick, tick, ...., tick". You can tell that >>I got most of my hardware education by listening over the wall. >> >Ah. Misunderstanding. The clock I'm talking about has nothing to do >with what time it is. Oh, I'm not talking about wall clock time. > ...It's the signal that keeps the computer in step >with itself. Right. And the base count was different depending on the power. Here in the States, the tick,tick was based on 60 Hz. In Europe, the tick,tick had a base of 50 Hz. > The real-time clock is an independent gadget that works >much like a digital wristwatch, but for some reason keeps worse time. On a timesharing system, where all operations had to be interruptable, keeping track of realtime was a bitch. That's why JMF heaved a sigh of relief when customer off-loaded that requirement to a mini. > >>> >>>There were certain advantages to this approach, but obviously not >>>enough. >> >>Would you explain one advantage for those of us who are still >>learning? >> >You have a series of gadgets, some fast and some slow. In an >asychronous system, each can run at its own best speed. If such a >system were clocked, the clock would have to be set for the slowest one. No. It wouldn't. The interrupt would just be set for a time interval, i.e., n ticks from now. Or a good design would allow the gadget a way to interrupt the system. That way the main processing wouldn't even have to worry about the gadget until the gadget raised its hand. The trick is to deal with all the raised hands in a fast, efficient, and accurate manner. > >Also, you don't have to spend money on a clock. There is that. When I began my system accounting project, we spent an awful lot of time thinking about time. (Somewhere, I even have a memo from a customer that talked about our masochistic invention dubbed Daylight Savings Time. But that's whole other thread topic.) We must have spent a couple weeks talking about the consequences of the operator entering the wrong date and/or time at a system startup. The obvious solution to all of our data gathering time stamp problems was to hook a teensy device that kept good time. A one-plus would give the device access to that place that keeps the date/time standard. I think the added manufacturing cost at that time was $100. It got nixed because the company didn't think having an accurate date/time on the system justified the added cost. [emoticon shaking its head at anal thinking]. Customers obviously cared or they wouldn't have been made such a big noise that started my project in the first place. I now regret that I didn't have a magic twanger that recorded all of those design discussions because there were reasons for the stuff we threw out and reasons for the stuff we kept. I keep getting the feeling that these exact same discussions are getting repeated. What a wierd feeling :-). /BAH Subtract a hundred and four for e-mail. ###### From: jmfbahciv@aol.com Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: Fri, 13 Oct 00 08:15:21 GMT Organization: UltraNet Communications, Inc. Lines: 30 Message-ID: <8s6rg4$jpf$2@bob.news.rcn.net> References: <8rm095$a4l$1@nnrp1.deja.com> <8s1h03$hif$5@bob.news.rcn.net> <8s2irf$5fi$1@panix2.panix.com> <8s46v0$9pf$3@bob.news.rcn.net> <8s4ofk$9ip$1@panix2.panix.com> <39e66029$3$wg$mr2ice@news.epix.net> X-Trace: +chL62/3K/O9jYAhdUgfkloCCxOpV7P+ZzXmAowSX00= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 13 Oct 2000 11:25:56 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!207-172-102-66 Xref: chonsp.franklin.ch alt.folklore.computers:65999 In article <39e66029$3$wg$mr2ice@news.epix.net>, jata@aepiax.net (Julian Thomas) wrote: >In <8s4ofk$9ip$1@panix2.panix.com>, on 10/12/00 > at 12:22 PM, phil@panix.com (Phil Gustafson) said: > >>Ah. Misunderstanding. The clock I'm talking about has nothing to do >>with what time it is. It's the signal that keeps the computer in step >>with itself. The real-time clock is an independent gadget that works >>much like a digital wristwatch, but for some reason keeps worse time. > >Yes - there's the cycle (or subcycle) clocking scheme, and there may be a >realtime or time-of-day clock. Real PITA sorting these two out when I >tried back in the late '80s doing a literature search on methods for >keeping RTCs in synch! > JMF was the guy who was in charge of the TOPS-10 monitor module that was essentially the device driver of the PDP-10 real time clock. I can't remember if he managed to deal with the device hung off the slave CPU in a master/slave configuration. It seems that was the way they were trying to configure it so that the master CPU couldn't be dragged away from the timesharing duties of the OS. I do remember that the real time clock interrupt had to take precedence over everything else since that data had to be captured RIGHT NOW and not a clock tick later. IIRC, the monitor module was called RTTRP.MAC. /BAH Subtract a hundred and four for e-mail. ###### From: phil@panix.com (Phil Gustafson) Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: 12 Oct 2000 12:22:12 -0400 Organization: Once-famed Parquet Floor Lines: 50 Message-ID: <8s4ofk$9ip$1@panix2.panix.com> References: <8rm095$a4l$1@nnrp1.deja.com> <8s1h03$hif$5@bob.news.rcn.net> <8s2irf$5fi$1@panix2.panix.com> <8s46v0$9pf$3@bob.news.rcn.net> NNTP-Posting-Host: panix2.panix.com X-Trace: news.panix.com 971367732 20892 166.84.0.227 (12 Oct 2000 16:22:12 GMT) X-Complaints-To: abuse@panix.com NNTP-Posting-Date: 12 Oct 2000 16:22:12 GMT 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!howland.erols.net!panix!news.panix.com!panix2.panix.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:66002 jmfbahciv@aol.com writes: >In article <8s2irf$5fi$1@panix2.panix.com>, >> >>Most computers these days are synchronous. There's a clock that counts >>cadence for the whole cpu. Every few nanoseconds, there comes a pulse >>that initiates a state change in the system. This state change is >>(usually) known to be complete before the next pulse comes along to >>initiate the next change. >> >>The KA-10 had no such clock. When some part of the system changed >>state, it sent a signal to the next part. The signal said, "Hey, I have >>some data for you on my output bus". The next part accepted the data >>and sent back an acknowledment: "Thanks for the data. You may go about >>your business". > >When you whisper these words in my ear, I immediately think about >hardware interrupts. But that's not what you are talking about... >or is it? Anyway, something had to keep a count of the ticks. >Wouldn't that have been called the clock? [red-cheeked emoticon >hangs its head, having just figured out it's reached a conclusion >based on associating a noun with a behaviour erroneously]. In my >non-guy head, an accumulation of tick counts meant that a clock >was going "tick, tick, tick, ...., tick". You can tell that >I got most of my hardware education by listening over the wall. > Ah. Misunderstanding. The clock I'm talking about has nothing to do with what time it is. It's the signal that keeps the computer in step with itself. The real-time clock is an independent gadget that works much like a digital wristwatch, but for some reason keeps worse time. >> >>There were certain advantages to this approach, but obviously not >>enough. > >Would you explain one advantage for those of us who are still >learning? > You have a series of gadgets, some fast and some slow. In an asychronous system, each can run at its own best speed. If such a system were clocked, the clock would have to be set for the slowest one. Also, you don't have to spend money on a clock. Phil -- )) (( Phil Gustafson Urban Legends FAQ: http://www.urbanlegends.com C|~~| Java FAQ: http://www.afu.com `--' ###### Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers References: <8rm095$a4l$1@nnrp1.deja.com> <8s1h03$hif$5@bob.news.rcn.net> <8s2irf$5fi$1@panix2.panix.com> <8s46v0$9pf$3@bob.news.rcn.net> <8s4ofk$9ip$1@panix2.panix.com> From: Ric Werme X-Newsreader: NN version 6.5.0 CURRENT #119 Lines: 36 Message-ID: <%btF5.42119$pu4.4417427@typhoon.ne.mediaone.net> Date: Fri, 13 Oct 2000 01:03:55 GMT NNTP-Posting-Host: 24.91.12.32 X-Complaints-To: abuse@mediaone.net X-Trace: typhoon.ne.mediaone.net 971399035 24.91.12.32 (Thu, 12 Oct 2000 21:03:55 EDT) NNTP-Posting-Date: Thu, 12 Oct 2000 21:03:55 EDT Organization: Road Runner Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!sunqbc.risq.qc.ca!cpk-news-hub1.bbnplanet.com!news.gtei.net!newsfeed.skycache.com!Cidera!cyclone-0.nyroc.rr.com!chnws02.mediaone.net!chnws05.ne.mediaone.net!24.128.8.70!typhoon.ne.mediaone.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:66093 phil@panix.com (Phil Gustafson) writes: >jmfbahciv@aol.com writes: >>When you whisper these words in my ear, I immediately think about >>hardware interrupts. But that's not what you are talking about... >>or is it? Anyway, something had to keep a count of the ticks. >>Wouldn't that have been called the clock? >Ah. Misunderstanding. The clock I'm talking about has nothing to do >with what time it is. It's the signal that keeps the computer in step >with itself. The real-time clock is an independent gadget that works >much like a digital wristwatch, but for some reason keeps worse time. While there was a separate real-time clock option for the -10 and places like C-MU implemented their own, the clock Barb is thinking of was the line clock. In the US, 60 hz, elsewhere 50 hz. Quite accurate, due to the way power distribution has to work! The clock Phil is talking about is the clock the steps a computer each "quantum" of the way. On the -10, this sort of clock would tick several times during a shift instruction, once for each shift. On the KA-10, there were several things call "monostables" or "one-shots" where you'd goose them with a pulse going in and they'd respond with a long pulse coming out. They timed a lot of things from the shift rate to the delay given on the I/O bus between asking for data and reading the bus. Not too many things had the handshaking Phil describe, IIRC. The memory system was the key one and allowed mixing memory of different speeds together. -Ric Werme -- Ric Werme | werme@nospam.mediaone.net http://people.ne.mediaone.net/werme | ^^^^^^^ delete Vote www.harrybrowne.com for President! ###### Newsgroups: alt.folklore.computers References: <8rm095$a4l$1@nnrp1.deja.com> <8s1h03$hif$5@bob.news.rcn.net> <8s2irf$5fi$1@panix2.panix.com> <8s46v0$9pf$3@bob.news.rcn.net> <8s4ofk$9ip$1@panix2.panix.com> From: jata@aepiax.net (Julian Thomas) Subject: Re: Lights Was: Tubes in computers Message-ID: <39e66029$3$wg$mr2ice@news.epix.net> X-Newsreader: MR/2 Internet Cruiser Edition for OS/2 v2.2a/20 Lines: 27 Date: Fri, 13 Oct 2000 01:06:52 GMT NNTP-Posting-Host: 199.224.125.91 X-Complaints-To: abuse@epix.net X-Trace: news1.epix.net 971399212 199.224.125.91 (Thu, 12 Oct 2000 21:06:52 EDT) NNTP-Posting-Date: Thu, 12 Oct 2000 21:06:52 EDT 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!nntp-out.monmouth.com!newspeer.monmouth.com!news2.epix.net!news1.epix.net!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:66030 In <8s4ofk$9ip$1@panix2.panix.com>, on 10/12/00 at 12:22 PM, phil@panix.com (Phil Gustafson) said: >Ah. Misunderstanding. The clock I'm talking about has nothing to do >with what time it is. It's the signal that keeps the computer in step >with itself. The real-time clock is an independent gadget that works >much like a digital wristwatch, but for some reason keeps worse time. Yes - there's the cycle (or subcycle) clocking scheme, and there may be a realtime or time-of-day clock. Real PITA sorting these two out when I tried back in the late '80s doing a literature search on methods for keeping RTCs in synch! -- Julian Thomas: jt . epix @ net http://home.epix.net/~jt remove letter a for email (or switch . and @) In the beautiful Finger Lakes Wine Country of New York State! Boardmember of POSSI.org - Phoenix OS/2 Society, Inc http://www.possi.org -- -- With sufficient thrust, pigs fly just fine. However, this is not necessarily a good idea. It is hard to be sure where they are going to land, and it could be dangerous sitting under them as they fly overhead. -Request for Comments: 1925 IOOF ###### From: phil@panix.com (Phil Gustafson) Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers Date: 12 Oct 2000 22:35:44 -0400 Organization: Once-famed Parquet Floor Lines: 28 Message-ID: <8s5se0$nps$1@panix2.panix.com> References: <8rm095$a4l$1@nnrp1.deja.com> <8s46v0$9pf$3@bob.news.rcn.net> <8s4ofk$9ip$1@panix2.panix.com> <%btF5.42119$pu4.4417427@typhoon.ne.mediaone.net> NNTP-Posting-Host: panix2.panix.com X-Trace: news.panix.com 971404544 1393 166.84.0.227 (13 Oct 2000 02:35:44 GMT) X-Complaints-To: abuse@panix.com NNTP-Posting-Date: 13 Oct 2000 02:35:44 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-stu1.dfn.de!news-koe1.dfn.de!news-was.dfn.de!znr.news.ans.net!cpk-news-hub1.bbnplanet.com!news.gtei.net!howland.erols.net!panix!news.panix.com!panix2.panix.com!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:66005 Ric Werme writes: > [Several things, but I'll focus on one.] > >On the KA-10, >there were several things call "monostables" or "one-shots" where you'd >goose them with a pulse going in and they'd respond with a long pulse >coming out. They timed a lot of things from the shift rate to the delay >given on the I/O bus between asking for data and reading the bus. > On an Unfortunate Hiatus from college I found myself applying regularly for technician jobs. Almost always, first thing, the interviewer would pass me a piece of paper and say "Draw me a schematic for a one-shot". One KA-10 one-shot had a timeout period just over 1/60 second, but never timed out because it was reset by a signal derived from, I think, a 6.3V filament transformer. If it wasn't reset, the power had failed, the interrupt got sent, and there was enough power in the capacitors to save the system state if you were lucky. I could draw a one-shot now, but not a Schmitt Trigger. Phil -- )) (( Phil Gustafson Urban Legends FAQ: http://www.urbanlegends.com C|~~| Java FAQ: http://www.afu.com `--' ###### Newsgroups: alt.folklore.computers Subject: Re: Lights Was: Tubes in computers References: <8rm095$a4l$1@nnrp1.deja.com> <8s1h03$hif$5@bob.news.rcn.net> <8s2irf$5fi$1@panix2.panix.com> <8s46v0$9pf$3@bob.news.rcn.net> <8s4ofk$9ip$1@panix2.panix.com> <%btF5.42119$pu4.4417427@typhoon.ne.mediaone.net> <8s6s00$jpf$3@bob.news.rcn.net> From: Ric Werme X-Newsreader: NN version 6.5.0 CURRENT #119 Lines: 95 Message-ID: Date: Sun, 15 Oct 2000 22:03:03 GMT NNTP-Posting-Host: 24.91.12.32 X-Complaints-To: abuse@mediaone.net X-Trace: typhoon.ne.mediaone.net 971647383 24.91.12.32 (Sun, 15 Oct 2000 18:03:03 EDT) NNTP-Posting-Date: Sun, 15 Oct 2000 18:03:03 EDT Organization: Road Runner Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.nextra.ch!news1.sunrise.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.cwix.com!chnws02.mediaone.net!chnws05.ne.mediaone.net!24.128.8.70!typhoon.ne.mediaone.net.POSTED!not-for-mail Xref: chonsp.franklin.ch alt.folklore.computers:66202 jmfbahciv@aol.com writes: >In article <%btF5.42119$pu4.4417427@typhoon.ne.mediaone.net>, > Ric Werme wrote: >>While there was a separate real-time clock option for the -10 and places >>like C-MU implemented their own, the clock Barb is thinking of was the >>line clock. In the US, 60 hz, elsewhere 50 hz. Quite accurate, due >>to the way power distribution has to work! >That's it!!!! What a difference a name makes :-) Thanks. I meant to add that the time between line clock interupts was called a "jiffy," you probably remember that in a TOPS-10 context. For example, the scheduler doled out CPU time in jiffies. E.g. if you had several compute bound jobs, each would run 10? jiffies and then some other job was scheduled. Hmm. 10 jiffies sounds too short from what I remember on console lights. >> >>The clock Phil is talking about is the clock the steps a computer each >>"quantum" of the way. On the -10, this sort of clock would tick several >>times during a shift instruction, once for each shift. On the KA-10, >>there were several things call "monostables" or "one-shots" where you'd >>goose them with a pulse going in and they'd respond with a long pulse >>coming out. They timed a lot of things from the shift rate to the delay >>given on the I/O bus between asking for data and reading the bus. >Whew! There's more to that hardware game than I had imagined. No >wonder it was "guy thing" ;-). >Weren't those pulses based on the line clock? No, we're talking orders of magnitude difference. The US jiffy, 1/60 sec is 16,667 microseconds, the time between new states in the CPU was mostly less than a microsecond. Or did a CPU designer >add a wire long enough so that it "timed" the speed of an electron >through it. That would get awfully expensive if you had to have >one/instruction and the timing was different for each instruction. That's actually done is certain applications, but usually not in CPU design. monstables generally use an "RC" (resistor-capacitor) network. A good mechanical analogy is blowing up a water balloon. The faucet is a variable resistor, the time from turning on the water and the balloon bursting is the "period" of the monostable. A resistor impedes electrical current flow, a capacitor stores a "charge" by collecting current flow. Instead of exploding, another circuit typically discharged the capacitor via a lower value resistor (i.e. qickly discharged), and stopped the current source. So the monostable was ready for it's next use. On the KL10 processor boards, the software that did layout did run etches back and forth to get the lengths the same. There was also a 2 inch-square zigzag that was "tapped" to distribute the clock signal. >>Not too many things had the handshaking Phil describe, IIRC. The memory >>system was the key one and allowed mixing memory of different speeds >>together. >Oh!!!!! [ emoticon here noticing that a light bulb just turned >on]. Is that what all of that meant. Didn't we allow mixes >but strongly recommended "don't do that"? Beats me. At C-MU we had MB10s (1.65 usec access time). When we got the free RCA prototype core box (1.0 usec access time), we were not about to shutoff the MB10s! So we put the MB10s in low memory and RCA in high memory where user programs ran. IIRC, there was some >talk about how the ordering of the boxes was a Very Important >Thing. Don't recall. Perhaps an issue with memory interleaving, certainly an issue with the wide memory access the KL10 did, IIRC. >Now, I sorta think I know how TW did his disk and magtape timing, >but would JMF have had to know the timing of the memory cycles >in order to do his memory management and CPU drivers? I don't think I knew what stuff he was measuring. Memory management generally doesn't have timing issues, CPU drivers generally have to deal with configuring the hardware, talking to DTE-20s, stuff like that. I spend a lot of time timing I/O rates and code paths. For the former, the system hardware clock is usually good enough (1,024 ticks per second), for code timing I use the Alpha CPU's cycle counter which ticks once per CPU clock, e.g. 500,000,000 times per second. It's remarkable what you can do in a microsecond these days. :-) -- Ric Werme | werme@nospam.mediaone.net http://people.ne.mediaone.net/werme | ^^^^^^^ delete Vote www.harrybrowne.com for President!