VLIW pre-history
|
From: Jan Vorbrüggen on 21 May 2007 03:23 > current day natural gas appears to have quite a bit of pipelinning > .... including multiple day latency ... lots of weather modeling. if they > don't start pumping in more gas, days before a cold snap ... they can > have inadequate gas at the consumer end. if they are wrong and the cold > snap doesn't occur ... they can have (over) pressure problems at the > consumer end (infrastructure with pipelines stretching hundreds of miles > from origin to destination). Make that thousands of miles, with several alternate routes. And you get another analogy: Because of the issues you mention above, there are buffers built into the pipeline, both at intermediate stations (a little) and near the end (a lot). In Germany, I believe these buffers could ride out a 90 day interruption of supply. Yeah, that's a lot of gas to store. But, as another poster noted, the stuff's compressible, and underground salt caverns are _huge_. Jan
From: Eugene Miya on 21 May 2007 13:26 In article <pan.2007.05.20.01.17.43.256323(a)stanfordalumni.org>, Leif Harcke <lharcke(a)stanfordalumni.org> wrote: >> some place with some obsolete obscure 120/164/264 processor attached to >> something like an MRI or CT will be found. > >The VAX 11/780 in the pedestal of the Goldstone tracking station had >one attached to its side up through 2002. It was part of the >real-time ranging system of the planetary radar. I believe it has >since been de-commissioned and surplussed. What the DSN replaced all those Modcomp IIs IVs and Classics with a VAX?! Which dish? the 264 ft. one or one of the smaller ones? Which model? I wonder if that used to be the 120B in 334. I'll raise the issue with Renfrow. --
From: Eugene Miya on 21 May 2007 13:32 In article <464e44a8.2446623(a)news.aioe.org>, John Savard <jsavard(a)excxn.aNOSPAMb.cdn.invalid> wrote: >>>"vaguely VLIW-ish", >>>dataflow > >I'm applying the term Dataflow to the Cyberplus from Control Data. I am not aware of any AFP/Cyberplus person who regarded it as a dataflow architecture. I see Bill Ragsdale annually, and I think it was he who gave me the AFP/Cyberplus manual that I had and eventually tossed (unfortunately before I started working with the Museum). Unfortunately, I'm skipping out of that conference this year. But I do on occasion see the younger Ragsdale, maybe next month and I will raise this. While I know the 205 203 and Star were regarded as super-CISCish I am not clear they regarded AFP processors as VLIW or dataflow. Alas should you find a physical instance of an AFP, then we can examine the issue in greater empirical detail. --
From: Eugene Miya on 21 May 2007 20:31 >Eugene Miya wrote: >> "In the beginning was the pipe. And the pipe was good. >> And Ken blessed the pipe..... > >> Pipelining goes back to 1964 in my biblio alone, but there are >> subtle terminology distnctions between all the different ways >> people use it (ask how errors propagate backward, that's a good first >> order question). In article <dlP3i.16047$p47.5679(a)bgtnsc04-news.ops.worldnet.att.net>, Stephen Fuld <S.Fuld(a)PleaseRemove.att.net> wrote: >I thought that pipelining went back at least to Richard Feynman's use of >pipelining his "computers", the human kind, in the work on the Manhattan >Project. :-) Functionally this is true, but one might also argue that this was an assembly line technique which goes back to Henry Ford and others. But that was merely part of an assembly line and in that way Nick is right that while the terminology might not have existed for pipelining people were smart enough back then to "look ahead" to do things asynchronously, etc. If this was a question of scheduling, the ACM Computer Surveys quarterly articles of the 70s on OSes cite management scheduling references. Then, trace scheduling would be no big deal. Why you guys might cite computing only references is up to you guys; but your faculty predecessors DID cite literature outside computing. From faulty memory: (examples) %A C. V. Ramamoorthy %A Hon F. Li %T Pipeline Architecture %J ACM Computing Surveys %V 9 %N 1 %D March 1977 %P 61-102 %K btartar, RBBRS1771, Ginsberg bib, computer architecture, pipelining, sequential processing, vector processing, CR categories: 5.24, 6.33 %X A good review including characterization, structure, pipeline hazards, branch and interrupt handling. A separate section covers vector processing. Many systems are covered. The Cray-1 and Amdahl 470/V6 are contrasted in detail. %X History, definitions of pipeline, and some mathematical analysis Asynchronous hazards (read after write, etc.). Pipeline Structure of the IBM 360/91. Details of Vector processing: TI ASC and CDC Star-100. Overview CRAY-1, and mention of the Amdahl 470 V/6. Reproduced in "Computer Architecture," D. D. Gajski, V. M. Milutinovic, H. J. Siegel, and B. P. Furht, eds., IEEE, '87, pp. 38-79. %A Mario J. Gonzalez, Jr. %T Deterministic processor scheduling %J Computing Surveys %V 9 %N 3 %D September 1977 %P 173-204 %K bsatya, RBBRS812, scheduling, frecommended91, dmp, ak, %X References are classified into various types (single, dual, multiple, and flow shop) at end of paper. --
From: Eugene Miya on 21 May 2007 20:45
>>> |> > Pipelining In article <f2r2nk$9ff$1(a)hubcap.clemson.edu>, Mark Smotherman <mark(a)clemson.edu> wrote: >other early pipelines: > >The Zuse Z3 (1941, patent filed in 1949) was pipelined: > > http://irb.cs.tu-berlin.de/~zuse/Konrad_Zuse/Z3-detail.htm I thought this was more a question on wide (horizontal) instruction systems than pipelining. >Gene Amdahl's WISC (1950) had a four-stage pipeline - instruction >fetch, operand fetch, execution, write back: > http://www.cs.wisc.edu/~bezenek/Stuff/amdahl_thesis.pdf So how would you compare a 4-stage pipe to say an 8-10 instruction word Multiflow Trace? >and the IBM Stretch and Univac LARC (projects started mid-1950s) >were pipelined. So do you think the Stretch was VLIW or a precursor? Finding those other references, I did find one on Leith who used the LARC for weather modeling. -- |