alanm@sunray.UUCP (Alan Myrvold) (06/05/89)
In article <6710021@hpcupt1.HP.COM> mount@hpcupt1.HP.COM (John Mount) writes: >>In article <d33P02nM30sj01@amdahl.uts.amdahl.com> shs@uts.amdahl.com >> (Steve Schoettler) writes: >>I guess I'm cheating but how about using tri-state logic and shifting >>right one tit? ;-). (Isn't tit the _accepted_ abbreviation for ternary digit?) >I always thought that tri-state logic's three states were HI, LOW >and HIGH_IMPEDANCE, the third state is used to take the chips outputs out >of the picture (cheap was to multiplex devices), I've never heard of anyone >building a device with three *signal* states. I recall hearing about a USSR computer that did base 3 arithmetic, and even had a Fortran compiler. Does anyone have details or a reference ??? - Alan --- Alan Myrvold 3755 Riverside Dr. uunet!mitel!sce!cognos!alanm Cognos Incorporated P.O. Box 9707 alanm@cognos.uucp (613) 738-1440 x5530 Ottawa, Ontario CANADA K1G 3Z4
nelson@berlioz (Ted Nelson) (06/07/89)
In article <6250@sunray.UUCP> alanm@cognos.UUCP (Alan Myrvold) writes: >I recall hearing about a USSR computer that did base 3 arithmetic, and even >had a Fortran compiler. Does anyone have details or a reference ??? Well, I know that their entire (?) first series at UNIVAC-type machines was base 3. Looking back at my notes, they had one called the SETUN which had an accuracy of 7.74E-9, but that's all the definite info I have. -- Ted.
news@ism780c.isc.com (News system) (06/08/89)
In article <6250@sunray.UUCP> alanm@cognos.UUCP (Alan Myrvold) writes: >I recall hearing about a USSR computer that did base 3 arithmetic, and even >had a Fortran compiler. Does anyone have details or a reference ??? > In the mid to late 50's I heard a lecture by Willis Ware, who had returned from a tour of the USSR's computer facilities. He did say that they had built a base three computer. But since they could not build tri-stable devices, they simulated them with a pair of flip flops. At the time of the lecture I am sure there was no FORTRAN compiler, because the only i/o devices were numeric. So there was no way to do FORTRAN. Also magnetic tape technology was so poor that a tape could be read only on the same unit that wrote it. Marv Rubinstein
baum@Apple.COM (Allen J. Baum) (06/09/89)
[] >In article <28301@ism780c.isc.com> marv@ism780.UUCP (Marvin Rubenstein) writes: >In article <6250@sunray.UUCP> alanm@cognos.UUCP (Alan Myrvold) writes: >>I recall hearing about a USSR computer that did base 3 arithmetic, and even >>had a Fortran compiler. Does anyone have details or a reference ??? >> > >In the mid to late 50's I heard a lecture by Willis Ware, who had returned >from a tour of the USSR's computer facilities. He did say that they had >built a base three computer. But since they could not build tri-stable >devices, they simulated them with a pair of flip flops. Hmm, I was under the impression that they actually built them with real three state device, called parametrons. This is transformer coupled logic, where the value on a signal is phase encoded, so 0 degree phase shift is 0, 120 degree phase shift is a one, 240 (or -120) degree phase shift is a 2. You can get all sort of logic functions, just be coupling the signals in strange ways with a transformer. -- baum@apple.com (408)974-3385 {decwrl,hplabs}!amdahl!apple!baum
desnoyer@Apple.COM (Peter Desnoyers) (06/10/89)
In article <6250@sunray.UUCP> alanm@cognos.UUCP (Alan Myrvold) writes: >I recall hearing about a USSR computer that did base 3 arithmetic, and even >had a Fortran compiler. Does anyone have details or a reference ??? > I seem to remember from a computer architecture course that using bases other than 2 is standard practice in constructing hardware multipliers, in order to limit carry propagation. Evidently you use base 3, with two wires per digit. Or something like that. On a different line, multi-level logic is used for some high-density ROMs. There was an issue of Spectrum (I think - or was it Computer?) on the subject about a year ago. Peter Desnoyers
pj@hrc63.co.uk (Mr P Johnson "Baddow") (06/26/89)
Nearest I ever heard of to this was in Heinlein's "The Number of the Beast", where there was a computer which used trinary. In a throw-away line, one character deduced that it must use three phase power. Ever since I read that I have been trying to figure out how it could work. -- Paul Johnson, | `The moving finger writes, And having writ, moves on,' GEC-Marconi Research | Omar Kyham when contemplating `vi'. ------------------------------------------------------------------------------ The company has put a radio inside my head: it controls everything I say!
ken@aiai.ed.ac.uk (Ken Johnson) (06/27/89)
In article <626@hrc63.co.uk> pj@hrc63.co.uk (Mr P Johnson "Baddow") writes: > >I have been trying to figure out how it [a base-3 computer] could work. One way would be to operate it to base 3 but have three digits +1, 0 and -1. Thus: +1 -1 -1 means 5 (+1*9 + -1*3 + -1*1) +1 +1 means 4 +1 0 means 3 +1 -1 means 2 (+1*3 + -1*1) +1 means 1 0 means 0 -1 means -1 -1 +1 means -2 (-1*3 + +1*1) -1 0 means -3 -1 -1 means -4 -1 +1 +1 means -5 (-1*9 + +1*3 + +1*1) etc. etc. -- Ken Johnson, AI Applications Institute, 80 South Bridge, Edinburgh EH1 1HN E-mail ken@aiai.ed.ac.uk, phone 031-225 4464 extension 212 `I have read your article, Mr. Johnson, and I am no wiser than when I started.' -- `Possibly not, sir, but far better informed.'
shafer@drynix.dfrf.nasa.gov (06/29/89)
In article <626@hrc63.co.uk> pj@hrc63.co.uk (Mr P Johnson "Baddow") writes: >Nearest I ever heard of to this was in Heinlein's "The Number of the Beast", >where there was a computer which used trinary. In a throw-away line, one >character deduced that it must use three phase power. When I was at UCLA, majoring in CS, one of my professors told me that the Russians had tried to build a trinary computer. The reason being that Shannon proved that e (2.7...) was the most efficient base for information content and 3 was closer to e than 2 was. Rather than using off-on, they used off-middle-high and for core memory, rather than just magnetized-nonmagnetized, they used clockwise, counter-clockwise, and nonmagnetized. He said that the theory was fine, but the implementation just killed them. It's easy to sense off or on, but a lot harder to sense off, medium, high. He claimed it set them back 10-15 years. I remember seeing this discussed somewhere else (IEEE, ACM?). -- M F Shafer |Ignore the reply-to address NASA Ames-Dryden Flight Research Facility |Use shafer@elxsi.dfrf.nasa.gov NASA management doesn't know what I'm doing and I don't know what they're doing, and everybody's happy this way.
entropy@pawl.rpi.edu (Math Student from Hell) (07/01/89)
In article <SHAFER.89Jun28150955@drynix.dfrf.nasa.gov> shafer@drynix.dfrf.nasa.gov writes: >When I was at UCLA, majoring in CS, one of my professors told me that >the Russians had tried to build a trinary computer. The reason being >that Shannon proved that e (2.7...) was the most efficient base for >information content and 3 was closer to e than 2 was. I've heard this before. Can someone in netland tell me what it means? What a wonderful thing is the human brain; how I wish I possessed one. Mark-Jason Dominus entropy@pawl.rpi.EDU entropy@rpitsmts (BITnet)
jk3k+@andrew.cmu.edu (Joe Keane) (07/02/89)
In article <5787@rpi.edu> puswad@pawl.rpi.edu (Math Student from Hell) writes: >In article <SHAFER.89Jun28150955@drynix.dfrf.nasa.gov> >shafer@drynix.dfrf.nasa.gov writes: >>When I was at UCLA, majoring in CS, one of my professors told me that >>the Russians had tried to build a trinary computer. The reason being >>that Shannon proved that e (2.7...) was the most efficient base for >>information content and 3 was closer to e than 2 was. > >I've heard this before. Can someone in netland tell me >what it means? _If_ the cost of a base-b digit is proportional to b, then the cost of something with N possibilities in base b is b*log_b(N), which is minimized at b=e, or b=3 if you restrict b to integers. Of course storing bits is actually _much_ easier than trits, so you have to take into account the value of `much'.
douyou@auto-trol.UUCP (Doug Young) (07/11/89)
In article <548@skye.ed.ac.uk>, ken@aiai.ed.ac.uk (Ken Johnson) writes: > In article <626@hrc63.co.uk> pj@hrc63.co.uk (Mr P Johnson "Baddow") writes: > > > >I have been trying to figure out how it [a base-3 computer] could work. > > One way would be to operate it to base 3 but have three digits > +1, 0 and -1. It is handy to use the characters P, M, and Z to represent +1, -1, and zero, respectively. Also, note that taking the "3's complement" of a number involves simply replacing P with M, M with P, and leaving Z as Z. For example PMZ ( (+1)9 + (-1)3 + (0)1 = 6 ) yields MPZ ( (-1)9 + (+1)3 + (0)1 = -6 ). To the digital mind, this may seem awkward. But consider circuitry which has three voltage levels: below a certain negative threshold is M, above a certain positive threshold (the same as TTL, perhaps) is P, and in between is interpreted as Z. In fact, one company is actually developing base 3 ("ternary"? "trinary"?) chips. It turns out you can avoid the carry-lookahead problem that plagues binary logic by going to tri-state logic. The new problem, however, is efficient conversion between the two representations. You win a few, you lose a few. Douglas Young ico!auto-trol!douyou Auto-trol Technology 12500 N. Washington Denver, CO 80241 (303)252-2418