Date: Tue, 20 Aug 1996 00:18:20 -0700 (PDT) From: Mark Crispin To: js-cgi@inwap.com Subject: PDP-10 web page Attached are a few files which you don't seem to have yet: Software Wars -- written 19 years ago, hopefully the statute of limitations has expired... Tony in RH20 Land -- an absolute must!!! PDP Summary -- all about various PDP's Mr. Bill Show -- Saturday Night Live meets DEC field service... -- Mark -- DoD #0105, R90/6 pilot FAX: (206) 842-0758 ICBM: N 47.36'24" W 122.34'08" TOPS-20: A Great Improvement Over Its Successors ================================================================== 17-May-91 21:28:54-LCL,7439;000000000000 Received: from hanna.cac.washington.edu by akbar.cac.washington.edu (5.65/UW-NDC Revision: 2.21 ) id AA07956; Fri, 17 May 91 21:28:53 -0700 Received: from life.ai.mit.edu by hanna.cac.washington.edu (5.65/UW-NDC Revision: 2.21 ) id AA23333; Fri, 17 May 91 21:28:48 -0700 Received: from Shiva.COM by life.ai.mit.edu (4.1/AI-4.10) id AA22260; Fri, 17 May 91 23:48:14 EDT Received: from Rosebud.Shiva.COM by Shiva.COM (1.34b) Fri, 17 May 91 23:50:53 EDT From: Phil Budne Received: by Rosebud.Shiva.COM (Spike-2.0) Fri, 17 May 91 23:48:17 EDT Date: Fri, 17 May 91 23:48:17 EDT Message-Id: <9105180348.AA12929@Rosebud.Shiva.COM> To: enders@plains.nodak.edu, pdp8-lovers@ai.mit.edu Subject: Re: PDP9, Anyone seen one? > Up to 32k words of 18 bit memory (yes indeed, this was an 18 bit > machine!) As were the 1, 4, 7 and 15! When I worked at DEC on 36 bit software, I met a few '15 folks who seemed almost as sad as we were in the end! I think I saw '7 in DEC salvage once. Bore a resemblance to a PDP-6 (dull blue with metal trim). I'm enclosing Mark Crispin's excellent PDP summary, since it hasn't been re-posted in a while! ================ >From RS%AI.ai.mit.edu@mc.lcs.mit.edu Thu Jun 8 00:28:01 1989 Date: Thu, 8 Jun 89 00:09:28 EDT From: "Robert E. Seastrom" Subject: What's a PDP-x ?? To: pdp8-lovers@mc.lcs.mit.edu If you folks have already seen this, please disregard it. If you haven't, keep it handy - it could come in handy some day. ---Rob PS: In case you didn't know (I didn't until 2 years ago), PDP stands for Programmed Data Processor... Date: Wednesday, 20 August 1986 03:42-EDT From: Mark Crispin To: TOPS-20@SU-SCORE.ARPA, Boken@RED.RUTGERS.EDU Re: DEC's PDP's Postal-Address: 1802 Hackett Ave.; Mountain View, CA 94043-4431 Phone: +1 (415) 968-1052 A number of people have requested my list of all the DEC PDP's, so I thought I'd bore you all with it. The PDP-1 was an 18 bit machine. It was DEC's first computer, and some of the first timesharing systems were designed for it. It's also unique in being one's complement; all later DEC computers were two's complement. Some machines, such as one of MIT's PDP-1s, were in operation until the late '70s. The PDP-2 was a designation reserved for a 24 bit machine, but as far as I can tell it was never even designed and definitely none were ever built. The PDP-3 was a 36 bit machine that was designed but never built by DEC. However, Scientific Engineering Institute built one in 1960. The PDP-4 was an 18 bit machine that was intended to be a cheaper, slower alternative to the PDP-1. It was so slow that it didn't sell well, although it was interesting for its auto-incrementing memory registers. It was not program-compatible with the PDP-1, but its instruction set was the basis of DEC's future 18 bit computers. The PDP-5 was a 12 bit machine designed to be a small laboratory system. It used many of the ideas in the LINC (Laboratory Instruction Computer, designed by Lincoln Labs at MIT, some of which were built by DEC). The PDP-6 was a 36 bit machine and the first machine to implement the most wonderful computer architecture known to man. It was rather expensive and difficult to maintain, and not many were sold. As a result, DEC cancelled 36 bit computers for what was to be the first of many times. The PDP-7 was an 18 bit machine and the sucessor to the PDP-4. It was a major price/performance win over the PDP-4 and the first DEC computer to use wire-wrapping. The PDP-8 was a 12 bit machine and the sucessor to the PDP-8. It basically defined the term "minicomputer", and went through several incarnations. The original PDP-8 was followed by the extremely slow PDP-8/S (as bad as the PDP-4 was to the PDP-1, but at least the /S was program-compatible). DEC recouped with the PDP-8/I (using MSI integrated circuits) and the smaller PDP-8/L, and somewhat later came out with the "Omnibus 8" machines -- the PDP-8/E, the PDP-8/F (a half-sized version of the PDP-8/E), the PDP-8/M (an OEM version of the PDP-8/F), and the final machine, the single board PDP-8/A. The PDP-8/A still exists after a fashion as a current DEC product. The PDP-9 was an 18 bit machine and the sucessor to the PDP-7. It had a faster memory than the PDP-7 and was the first microprogrammed DEC computer. Modulo a 300 wire(!) ECO required in the first machines, the PDP-9 was a reliable machine and some are still in operation today. There was a short-lived PDP-9/L. The PDP-10 was a 36 bit machine and the sucessor to the PDP-6. It is especially noted for its software, which represents the pinnacle of DEC software engineering and has never been equalled. The first KA10, largely installed in universities, created a whole generation of timesharing hackers. The follow-on KI10, with paging and using IC's instead of discrete components but otherwise unexciting, mostly was sold to commercial organizations. The KL10 went through several incarnations and is today the most representative of this marvelous machine. The KS10 was a small, low-speed (approximately KA10 performance) processor which was DEC's last successful implementation of this architecture. The PDP-11 was a 16-bit machine that went through more implementations and operating systems than can be counted. Presently it superceded the less powerful PDP-8 as the representative minicomputer. While the PDP-11 used octal, it was in its deep heart of hearts a hexidecimal machine, and the first indicator of the creeping IBMification of DEC that took full fruit in the VAX. [I can hear the flames now...] Rather than fight it the customers loved it; more PDP-11's have been sold than any other DEC computer (possibly more than all the others combined). The PDP-12 was a 12 bit machine and the sucessor to the PDP-8. It combined a LINC and a PDP-8 type processor in the same box and basically was a new model of the LINC-8 which was the same thing. No PDP-13 was ever designed or built. Even DEC gets superstitious. The PDP-14 was a 12 bit machine with a 1 bit register. It was used as a process control engine in applications that were felt to be too rugged for a PDP-8, and basically replaced a set of relays. Later DEC made PDP-8's suitable for this sort of thing, but it didn't stop them from the ultimate silliness of building a PDP-14 that used a PDP-8 as its console processor! The PDP-15 was an 18 bit machine and the final one of this design built by DEC. More PDP-15's were built and sold than any of the others, and it went through several incarnations including some which used a PDP-11 as a front end. Apparently the cancellation of the PDP-15 came as a great surprise to the "Tiger Team" who worked on it, although considering its general ungainliness compared to comparable performance PDP-11's it wasn't surprising. In many ways the PDP-15 died for the same reason the PDP-10 did. The PDP-16 was a "roll your own" 16 bit machine based on various "building blocks". Every PDP-16 was essentially custom-designed by the customer. It got a fair amount of attention when it was announced but evidentally didn't sell very well. There was no PDP-17 or any other designator. DEC apparently decided that "PDP" had a perjorative ring to it. 22-May-91 04:19:30-LCL,17200;000000000000 Received: from hanna.cac.washington.edu by akbar.cac.washington.edu (5.65/UW-NDC Revision: 2.21 ) id AA08510; Wed, 22 May 91 04:19:29 -0700 Received: from life.ai.mit.edu by hanna.cac.washington.edu (5.65/UW-NDC Revision: 2.21 ) id AA02662; Wed, 22 May 91 04:19:24 -0700 Received: from watsun.cc.columbia.edu by life.ai.mit.edu (4.1/AI-4.10) id AA00689; Wed, 22 May 91 07:02:50 EDT Received: by watsun.cc.columbia.edu (5.59/FCB) id AA04188; Wed, 22 May 91 05:17:26 EDT Date: Wed, 22 May 91 5:17:25 EDT From: Charles Lasner To: pdp8-lovers@ai.mit.edu Message-Id: From: Charles Lasner Subj: Corrections and commentary on Mark Crispin's PDP list I have found "grievous" errors in Mark Crispin's descriptions of PDP-xx computers from DEC, so herewith are some corrections: PDP-1, 2, 3 I have no fundamental problem with this description, however, I would note that there is a story circulating that the original machine was called PDP, *not* PDP-1. The manual for the PDP was PDP-1, and they intended it to stop there. PDP-2 and PDP-3 are therefore tentative or actual notation for more manuals. I will not comment other than this mention of the PDP-2, 3 memorial golf course :-) PDP-4 I once saw one of these in the Mill with a baudot teletype (model 28) hooked up to it; no further comment. PDP-5 The PDP-5 is nearly compatible with the PDP-8. It has the program counter in location zero. Storing value-1 in 0000 is the same as a jump to value. This is how FOCAL, 1969 tests for running on the -5, since interrupts must be moved from 0001 to 0002. All other references to 0000 are disallowed, as are certain operate instructions. These restrictions seem arbitrary, and this was corrected in the PDP-8. It's probably not unreasonable to modify this machine to be -8 compatible, although a little slow due to 8 microsecond memory. Extended memory is totally compatible, but EAE is not, it's more like a peripheral than a processor option. The bus can't do 3-cycle data break, but the early 555 DecTape ran on it in 1-cycle mode using the current Data Field as the break field (which in turns means no interrupts are allowed while the DMA is in progress!). Other than that, through use of a physical bus convertor (don't have the number, but I mean a passive device that merely changes the connectors), the -5 likes -8 negative bus peripherals. An -8 expander box (804) was once hooked to a -5 containing the -8's plotter interface for a Calcomp plotter, etc. Rumor has it that someone modified the CPU of a -5 to gain the few operations necessary to make it -8 compatible, and that an 8K version was running OS/8 on some undisclosed peripheral. I wouldn't give too much creedence to the story because all of the early disk peripherals were 3-cycle except the aforementioned 555 DecTape (as opposed to TC01 et al which is three-cycle). The PDP-5 is clearly the PDP-8's father. It is not known if it used the PC in 0000 kludge as an economy measure, or just plain foolishness at the time, but clearly, had it been slightly upgraded to PDP-8 processor compatibility by DEC, it would have been supported more readily. Many of the changes subsequently made to FOCAL, 1969, etc. depend on removing restrictive operations caused by supporting the -5 and the even more demented 8/s (see below). These restrictions mean that you can't use combined operates such as CMA RAL because these two can't do it. In the -5's case, it's almost excusable, but in the -8/s case it is not. PDP-6 The story I heard is that Alan Kotok virtually single-handedly got all of the few made to work by using other people's memory. Most wound up in MIT at one time or another anyway. The PDP-6 lacked certain features required by later -10 systems, but was always welcome to ITS. PDP-7 It is interesting to note that a small class of peripherals were introduced for either the -7/9 series or the -8. I know of at least two: the AA01A for the -8, and the type 884 sychronous modem interface. Each was delivered in two parts. One part was bus specific, and the other was plugged into whichever first type you had installed appropriately. I personally participated in modifying the D-A part of a AA01A to become an -8-only peripheral, because the trivial -8 interface was a waste of valuable black block racks, and the empty space on the AA01A was available to wrap on an -8-only interface complete with daisy-chain. I also once used a PDP-15 with a negative bus converter which had the negative interface and 884 assembly for a synchronous modem connection to an outside dedicated line. It has been told to me that the TC58/59 class of tape controllers consists of a generic subsystem called TC50 internally, which is wrapped to a 7/9 bus or -8 bus specific portion to become the TC58 or TC59. In this case, the two are wrapped together, not cabled together, and are thus more or less permanently attached. PDP-8 and PDP-12 Since the PDP-8 is the standard bearer of our group, we can't leave any stone unturned here. First off, the obvious typo in that the -8 is the successor to the -5, not itself :=). While the 8/s may be extremely slow, it is also *not* compatible with the -8. Please note that OS/8 does *not* run on either the -5 or the 8/s. This also applies to P?S/8, which it should be noted runs on the entire -8 family from PDP-8 through DECmate III+, whereas OS/8 has different "family" members that each do part of this in an overlapping fashion. I am part of a group which desires to release OS/8 Version 5, to run on the entire family. Note, when all concerned parties discuss the "family of 8" the 8/s and -5 are specifically omitted. I have once discussed the feasibility of upgrading the 8/s, and the consensus was that it could be done, but it might run a bit slower, and in any case would require a rack of cards hanging out of it. It is also noteworthy that the 8/s was originally a different project known as PDP-10! It was to have 32K memory implemented as drum-style memory as in earlier computers, and the memory was what became the basis for the 32K -8 fixed disk peripheral called DF-32. I guess they didn't pay close attention to the improvements in processor OPR instructions the -8 made over the -5, because the 8/s is *worse* than the -5 in terms of compatibility. It is also a fully *serial* computer, not parallel. It should also be noted that many DATA GENERAL Nova-type machines and microes were serial or at least not fully parallel. Later models, the 8/i and 8/l did come out. Early 8/i's had Red-handled modules like the -8, even though the cards were TTL and eventually became maroon (M modules). There were even a few -8's with 8K in them ala the 8/i, because they had 8/i-type memory kludged into them! The 8/i originally had a negative bus, for compatibility with the vast majority of existing peripherals made expressly for the -8. The 8/l is essentially a stripped-down 8/i, and as such can have only 8 (or 12) K and a positive bus, whereas later 8/i's could have either bus. The LINC-8 was an attempt to bring DEC some income from the MIT LINC project. After early on realizing that an independent LINC peripheral would be difficult to pull off, the LINC-8 was born by making moderate modifications to the -8 CPU and memory interface, and adding a "captive" processor. Many operations had to be simulated for complete pseudo-compatibility with the "real: LINC, which in turn was constructed of the same System Modules as the PDP-5. The LINCtape peripheral is actually an -8 peripheral, whose major registers are the LINC processor's registers, which are "borrowed" because the LINC CPU is totally disabled while the -8 runs the tape peripheral. In certain ways, the LINC-8 is a "cadillac" implementation of the -8, in that it has certain niceties such as GMT 3 grasshopper fuses on all power busses, and many fans, etc. Also, the lamp drivers for all front panel lamps are on actual small modules wrapped in, not one massive (and damage-prone) lamp driver panel. The bulbs still are soldered in here though :-( The PDP-12 is a positive bus-only 8/i with a built-in DMA multiplexor option eventually, as well as push-down stack hardware and priority interrupt (KF-12B) as late-production standard features. All of the LINC stuff is "real" including the dual console. There are less actual lights than the LINC-8 because there is now only one set of registers for a dual-mode processor, not two complete processors with each's set of major registers displayed. A standard option allows read/write of DECtape, whereas only user modifications could do this on the LINC-8 eventually. The FPP-12 was born here, and included PDP-8 CPU lockout mode option, although this was reborn in the FPP-8/e,a for the Omnibus years later. Clearly, the -12 was the one model that got the benefit of "proper" hardware engineering, whereas subsequent models seem to get ever shorter ends of the hardware "stick". Curiously, a few modules come from the LINC-8 project, because it was too much work to recreate the function in TTL rather than negative voltage logic, so they merely level converted and used a few R-series cards :-). Some PDP-12 modules actually say they are specifics for the "LINC-8/i" which was apparently its original designation. The PDP-12 is the only model where the number is also the word length :-). Many DEC-ignorant computer types actually believe that the PDP-8 is an 8-bit machine! Newer models of -8's are all for the Omnibus, which defines all processor and I/O signals on four DEC-feet or quad-width slot. Various incarnations are either wider or deeper, but all have the same bus. The standard is the 8/e, which is always 20 slots (optionally 40) in the full-depth box, complete with bulky analog power supply. The entry level is the 8/m, which was designated for OEMs, and didn't include the standard 8/e-type front panel with incandescant lamps. The 8/f is an 8/m including an LED variant of the lamp panel, and this panel is an option for the 8/m as well. Thus, the 8/m with panel and 8/f differ only in color scheme and silk-screened name on the panel. Both have a slightly balky switching supply which is much smaller than the 8/e supply, as well as much less power capability. These machines are "bread-box" size, and only allow for 20 slots in the box. The newer breed of Omnibus boxes are all usually known collectively as 8/a, but they came in various flavors: a short-lived 10-slot box with many restrictions, a 12-slot box with somewhat more versatility, and a 20-slot box with large power capability. Theoretically, any of these systems could be expanded from virtually any model's box as an add-on to the base system, so standard options included upgrading a 40-slot to 80-slot, or making a 20-slot 8/f grow to 40 or optionally 60 slots using the same external 8/e-type box, etc. The FPP-8/A was meant to go in the 8/a 12-slot box, which like all 8/a boxes is hex-wide internally, not quad. The sixth foot is non-existent and the fifth foot has only a few signals optionally available later to various memory extension options beyond 32K. When sold this way as an add-on for a 40-slot 8/e, the entire option is known as FPP-8/e, including the BC-80C cable and 12-slot hex-wide box, as well as the two hex-wide FPP cards. My own machine is a 40-slot 8/e, with a 20-slot 8/a box as an expander. The 8/a series also sports an optional octal digit LED readout front panel, similar to an 11/34 in general appearance. Most of the 8/a models use a single slot CPU, the KK8A, which has the bus pullups on it (at the *wrong* end electrically of the bus!) which disallows the bus terminator, and is thus limited to what fits in any one box, or at most 20 slots. Some use the 8/e processor (KK8F) which requires a terminator card (M8320) in the furthest slot away from the CPU. Some configurations reverse the apparent beginning and ending, but always obey this rule. This allows at least 40 slots (8/e box or two 8/a boxes) or at most 80 slots (two 8/e boxes). It is possible to discover which CPU is present due to specific CPU-model-dependent quirks, but you can't tell more than that. Some 8/a boxes are really 8/e machines with 8/a memory options and all else is from an 8/e! 8/a systems get numeric suffixes like -100, -400, -500, and -600 as well as model-option changes to make it something like -625, etc. Since there is much "plug and play" here, it is not too useful to describe an 8/a system by model, but rather by specific components. All -8 models are backwards compatible with earlier bus arrangements, and also there is a feeble attempt at forwards compatibility called the DW-8/E. This partially allows some Omnibus stuff to work in the pre-Omnibus machines, but has many restrictions. A hacked-up version of the RK8E called RK8F is the only DMA option for the DW8E. The Omnibus machines support the positive bus via the KA8E and KD8E(s). The KA8E outputs a completely compatible positive transfer bus, and each KD8E outputs a compatible DMA interface. The Omnibus does any priority multiplexing. The Negative bus is in turn supported by thge DW08A option, so it is reasonable to fine a TC01 on an 8/a. I have an 8/e with a TC01 and TU55 myself. (This requires an ECO to the W103 cards to become W123.) PDP-9 and PDP-9/l The 9/l had a pink color scheme, and was apparently a stripped-down -9. The -9 was often sold with a -7 because there was some software that ran on both simultaneously. Apparently, the -7 became a dedicated controller for the DECtape drives relative to the -9, because the -7 had a feeble interface ala the TD8E on the 8/e, etc. Later, the -9 got its own TC02 DECtape, thus relieving the -7 requirement. I don't understand Mark Crispin's description of the -9 as being microprogrammed. Perhaps this is merely a misstatement of the fact that the operate instructions, like in the -8, can be *user* microprogrammed together such as: SMA SZA CLA /skip if the AC is either negative or zero, /and then clear the AC regardless. The -9 is not materially different in architecture from the -8, so I can't imagine this meaning anything else. I also understand that the 7/9/15 can microprogram more of these operations than the -8 can. The -8 is restricted to groups which can combine sub-operations, whereas the 18-bit machines can combine almost anything at once! PDP-10 The PDP-10 spawned much software development, and eventually encouraged the use of BLISS. It is noteworthy that much of the VAX software is lifted directly from TOPS-20 because of the close compatibility of the various (later) BLISS implementations. However, purists will note that BLISS-10 is *not* BLISS-36, while BLISS-32 just about is. This was done on purpose to rip off the -10 stuff for the benefit of the VAX. Perhaps this also means that some newer stuff can come back to the 36-bit world as well :-) PDP-11 Since the early days of the RISC discussion, it has become debatable whether the PDP-11 is more powerful than many designs, not just the -8. I guess it all depends on how you measure the power. The PDP-11 certainly had more hardware engineering effort applied, but was the benefit worth the cost? Measured this way, you could say the -11 is DEC's weakest machine. Look at how well the PDP-8 has fared while being "negatively marketed" as compared to the -11! Also, LCG had to put controls on -11 marketing when salesmen were quoting -11 systems that were a) impossible to implement (or at least highly impractical), and b) were clearly short-changing a customer willing to shell out PDP-10-sized bucks for a computer. DEC also made lots of money selling the 11-03, but I guess -11 freaks don't want to talk much about that :-) Various academic works have pinpointed the weaknesses of the -11 in terms of pure architectural considerations. The user group largely consists of people who don't really know much about the -11, but rather are in love with its software, and make the fatal mistake of measuring the worth of a machine by how useful it is to them, not its intrinsic value. This merely means that DEC threw a lot of money at the -11, not that it is a good design. If RT-11 were known as RT-15 for the PDP-15 family, and DEC implemented an RT-11-like system for the -15 (which is totally feasible, considering what a -15 is, and what RT-11 is), how many of these people would be raving about how powerful the -15 is? :-) As a side note, there seems to be much interest in an RT-11-like system for the VAX, presumably to be called RT-32. As is alluded to in THE SOUL OF A NEW MACHINE, the PDP-11 was born over a weekend, appealing to the ego of certain individuals who could pull strings in DEC, to pull the plug on what was to be known as DATA GENERAL and the NOVA. Considering the rocky road of the -11, it is curious that DEC never sued DATA GENERAL over the NOVA's very existence. cjl