Bootstrap ancestry
in [UK Mac]
|
From: Rowland McDonnell on 19 Jun 2010 08:29 Ian McCall <ian(a)eruvia.org> wrote: > real-not-anti-spam-address(a)apple-juice.co.uk (D.M. Procida) said: > > > You can do the same thing with computer technology; if it - but nothing > > else - were all wiped out, where would we go from there? > > Multicore multighz Spectrums for all! The People's Computer(tm). > > > Taking it all a bit more seriously, I wonder if von Neuman would > necessarily have won out. He didn't `win out' in computer designs, it's just that there was a sensible way to approach the job that was easy and practical and so on, and everyone did it the same way because of that. Yes, JvN did a lot of good work, but he's nothing like the father of electronic computers. Alan Turing has a much greater claim to that mantle if you ask me - not that any one person deserves it. > It's almost ingrained now that 'this is how > you design a computer', but von Neuman didn't think that. He thought > 'this is how you design a computer practically and economically'. > Competing designs existed, and for me it's interesting to see aspects > of them come back - clockless for example. Plenty of computer designers just did their own thing, totally independent of von Neumann. It's just that the US-centric view of world history likes to claim that the USA invented All Good Things, so presents JvN as the Primary Source for It All, when he was just one of the many players. "First Draft of a Report on the EDVAC" is presented as some kind of foundation stone that he created. It was more of a log of what had been worked out up to that point. Not so much the primary influence, but more a record of the influences to date. Asyncronous computing has always been with us, it's just that synchronous is easier to design. I used to hang out with one of the Amulet team at Manchester Uni... > Far back in the mists of time I remember being taught about the > Manchester Butterfly, something I curiously can find very little > (actually, nothing at all) about on the web. It was a forerunner to all > the multicore designs being used at the moment, but was abandoned > because the interconnects between its nodes were not fast enough at the > time to overcome the cost of signalling a task had been done. How old was that design, then? > The > conspiracy theory in me notes that the PS3's cell, itself semi-based on > Blue Gene, lists a company called Butterfly amongst its credits and the > PS3's/Blue Gene's designs are very Manchester Butterfly, at least as > far as I remember it being described to me at University. We'd also > have taken things like the Transputer a bit more seriously too. That one *did* have the high-speed interconnects required. I though we always did take the Transputer seriously? Who's the `We' whom you think didn't? Rowland. -- Remove the animal for email address: rowland.mcdonnell(a)dog.physics.org Sorry - the spam got to me http://www.mag-uk.org http://www.bmf.co.uk UK biker? Join MAG and the BMF and stop the Eurocrats banning biking
From: Rowland McDonnell on 19 Jun 2010 08:43 Richard Tobin <richard(a)cogsci.ed.ac.uk> wrote: > The newsreader I am using now was compiled several years ago on a PPC > Mac, using gcc. That gcc was no doubt compiled by Apple using another > copy of gcc they had compiled. A few generations back, they probably > cross-compiled gcc, perhaps on a 68000 Mac or Next computer. A few > generations before that, gcc was compiled on a Vax using the BSD C > compiler. That C compiler descended from Richie's original C compiler > - I'm not sure whether that was written in B or assmebler. If you > keep going back, you will eventually come to a program made, not > begotten, entered using switches with nothing but hardware behind > them. > > I could have followed many branches instead of just the compilers - > the editors, assemblers, linkers and operating system used along the > way, not to mention the computers they ran on. Every program has a > branching and converging bootstrap ancestry. Why say that? I don't see that's the case at all. > How many long-lost > programs still have running descendants? How many parentless programs > is the entire ecosystem of modern software descended from? How do you measure it? I don't think it's sensible to suggest that a program written from scratch without prior code in it is descended from other code, even if it's written using a compiler or whatnot. I don't see that your model is necessarily valid. Rowland. -- Remove the animal for email address: rowland.mcdonnell(a)dog.physics.org Sorry - the spam got to me http://www.mag-uk.org http://www.bmf.co.uk UK biker? Join MAG and the BMF and stop the Eurocrats banning biking
From: Justin C on 19 Jun 2010 08:18 In article <883cceFgicU1(a)mid.individual.net>, Chris Ridd wrote: > What sort of blind alleys would we avoid? x86? I think that knowing, in advance, that while at first 32bit was liberating, it became a mill-stone, and backward compatibility issues sure hindered progress. 64bit good, 128bit better, 256bit? Would we try and find a way to avoid this kind of thing completely? Massive parallelism? To get more bits add another cpu? I think that the architecture could be very different. This is the sort of question that it would be great to ask of those people who have worked 8bit 16bit 32bit and 64bit - if you could start again, how would you do it? Justin. -- Justin C, by the sea.
From: Justin C on 19 Jun 2010 08:22 In article <883cceFgicU1(a)mid.individual.net>, Chris Ridd wrote: > What sort of blind alleys would we avoid? Copper telephone lines to each house - we'd put fibre optic in from the outset! Justin. -- Justin C, by the sea.
From: Ian McCall on 19 Jun 2010 10:42
On 2010-06-19 13:29:32 +0100, real-address-in-sig(a)flur.bltigibbet.invalid (Rowland McDonnell) said: > Ian McCall <ian(a)eruvia.org> wrote: > >> ...Taking it all a bit more seriously, I wonder if von Neuman would >> necessarily have won out. > > He didn't `win out' in computer designs, it's just that there was a > sensible way to approach the job that was easy and practical and so on, > and everyone did it the same way because of that. Yes, that's what I meant. It's just that it has become thought of as a truism that computers Are Just Like That(tm), whereas as you say there were other designs around at the time. >> Far back in the mists of time I remember being taught about the >> Manchester Butterfly, something I curiously can find very little >> (actually, nothing at all) about on the web. It was a forerunner to all >> the multicore designs being used at the moment, but was abandoned >> because the interconnects between its nodes were not fast enough at the >> time to overcome the cost of signalling a task had been done. > > How old was that design, then? I wracking my brains, nut must admit I don't actually know if it was a 50s or 60s design. Information on it is curiously absent from the web, but I can clearly remember the lecture I was given at university on it. Always been looking to see if I could find something more concrete on it since. > >> The >> conspiracy theory in me notes that the PS3's cell, itself semi-based on >> Blue Gene, lists a company called Butterfly amongst its credits and the >> PS3's/Blue Gene's designs are very Manchester Butterfly, at least as >> far as I remember it being described to me at University. We'd also >> have taken things like the Transputer a bit more seriously too. > > That one *did* have the high-speed interconnects required. I though we > always did take the Transputer seriously? Who's the `We' whom you think > didn't? We being the general marketplace, not particular individuals within it. Transputers never really caught on in a big way, and all was said to be down to optimising compilers and languages for multithreading and getting developers to understand it properly. Of course, twenty/twenty five years later we don't have that problem of course, with all compilers languages having been fully optimised for multithreading and all developers fully able to take advantage of it without trouble. I mean, imagine someone in this day and age who just puts things on a single thread, or who doesn't understand fully how to safely code in a multithreaded, multicore environment. Ridiculous! Err....hmm. Cheers, Ian |