al@perisl.UUCP (Al Schuilenburg) (03/01/91)
In comp.sys.transputer <moses@nadc.navy.mil> writes > This brings up a simple question: Who wants massively parallel > machines anyhow? Computer Scientists and Engineers experiment > with and develop systems - what are they used for? I'm all > for research for its own sake, but there are actually people > out there who want to apply it. I suppose the questions are more > along these lines: What are the foreseen uses of large parallel > machines? Is this in line with current research? What should > they be used for (besides weather prediction, Mandelbrot sets, etc.)? > > Just wondering There are a large number of applications for large parallel machines which are ideally suited to parallelism. For example, Edinburough Parallel Computing centre has a directory about an inch thick of its current projects, with Southampton University also doing a fair amount (not to mention Cambridge University and a number of other institutions in the UK alone too numerous to mention) doing quite a bit of parallel applications, let alone in the parallel processing research field. This is because there are a number of applications which really chew up vast amounts of computing resources (e.g. Cellular Automata, Fluid Flow) where places like Shell Oil in Amsterdam have a 400 transputer machine to handle applications such as these where even your CRAYs would step back and think twice about doing. I suggest you try and contact some of these people if you genuinely are interested. Granted the transputer is popular among engineers and scientists who need more power than their PC can give, but ever thought that this is the one fairly cheap method whereby such researchers can get their hands on computing power at a cheaper price (we cant all afford CRAYs etc). Research always has been notoriously short of funding, and parallel processing provides the power at a cheaper price. Anyway, I feel that parallel processing is where the future of processing lies. Ever thought what would happen when the limits of nature were reached and single processors could be made to run no faster? Man's desire for faster, better and better would soon turn to "put more on the job". However, enough idle chatter..... Alex al@perisl.uucp Disclaimer: The views expressed in this article are my own and should not be taken as that of my workplace.
anand@wotan.top.cis.syr.edu (Rangachari Anand) (03/02/91)
In article <9103011038.AA15028@perisl.uucp> al@perisl.UUCP (Al Schuilenburg) writes: >lies. Ever thought what would happen when the limits of nature were >reached and single processors could be made to run no faster? Man's >desire for faster, better and better would soon turn to "put more on >the job". However, enough idle chatter..... Such statements are undoubtedly correct in the long run, but I point out that there seems to be no slowing down of increases in processor speeds. I draw your attention to the new RS6000 series, the unannounced HP machine (said to be a screamer) and the 100 MIPS SPARC processor from TI. The real danger in parallel machines is that they seem to get obsolete alarmingly fast. R. Anand anand@top.cis.syr.edu
ong@d.cs.okstate.edu (ONG ENG TENG) (03/03/91)
From article <9103011038.AA15028@perisl.uucp>, by al@perisl.UUCP (Al Schuilenburg): > In comp.sys.transputer <moses@nadc.navy.mil> writes > >> This brings up a simple question: Who wants massively parallel >> machines anyhow? Computer Scientists and Engineers experiment >> with and develop systems - what are they used for? I'm all >> for research for its own sake, but there are actually people >> out there who want to apply it. I suppose the questions are more >> along these lines: What are the foreseen uses of large parallel >> machines? Is this in line with current research? What should >> they be used for (besides weather prediction, Mandelbrot sets, etc.)? >> >> Just wondering I was a Phd intern at ARCO's research facility in Plano, TX last summer, and boy can they use massively parallel machine. One guy has a program that runs on a 6-processor IBM 3090 for weeks! And that was with a *small* dataset!
al@perisl.UUCP (Al Schuilenburg) (03/06/91)
In article <9103011038.AA15028@perisl.uucp> Rangachari Anand writes >In article <9103011038.AA15028@perisl.uucp> al@perisl.UUCP (Al Schuilenburg) writes: >>lies. Ever thought what would happen when the limits of nature were >>reached and single processors could be made to run no faster? Man's >>desire for faster, better and better would soon turn to "put more on >>the job". However, enough idle chatter..... > >Such statements are undoubtedly correct in the long run, but I point >out that there seems to be no slowing down of increases in processor >speeds. I draw your attention to the new RS6000 series, the unannounced >HP machine (said to be a screamer) and the 100 MIPS SPARC processor from >TI. The real danger in parallel machines is that they seem to get >obsolete alarmingly fast. With reference to the lack of slowing of processor speeds, imagine the power you would have with 256 (why not ?) or so of RS6000 series or the 100 MIPS SPARC processors working in parallel. A hell of a lot faster than one. Also I would disagree with you about parallel machines becoming obsolete alarmingly fast. Most of the resonably large parallel machines I know of are still in regular use. Maybe their individual processors are coming to the end of their sequential lives, but not their parallel ones. When this happens the prices of the parallel machines drop, making it cheaper to build bigger ones and more accessible to the researchers requiring the power. However, one must not forget Amdahl's Law............... Alex al@perisl.uucp