Author: Robert Hyatt
Date: 17:50:03 10/05/01
Go up one level in this thread
On October 05, 2001 at 13:32:09, Bruce Moreland wrote: >On October 04, 2001 at 18:42:17, Dave Gomboc wrote: > >>On October 04, 2001 at 14:54:01, Bruce Moreland wrote: >> >>>On October 04, 2001 at 06:46:48, Sune Fischer wrote: >>> >>>>On October 04, 2001 at 04:56:08, Uri Blass wrote: >>>> >>>>>The difference is that the other side(I and Bruce) do not say that they know >>>>>that there is a super linear improvement but that we cannot say that it is >>>>>impossible with the known algorithm of today. >>>>> >>>>>I said that the only way is to investigate the problem by test positions in >>>>>order to see if programs can get a super linear improvement from 2 processors. >>>> >>>>But that is not the only way, you can also use logic. >>>>This is like explaning to an inventor why he can't make a perpetuum mobile >>>>machine. If he doesn't understand the laws of physics, does not know of energy >>>>conservation, then he will keep arguing till the day he die, that "we can not >>>>know for certain until we have tried everything". >>>> >>>>We _do_ know, there is proof and Bob has outlined it several times, but if you >>>>won't listen or understand, then we have a communication problem. >>>> >>>>-S. >>> >>>A perpetual motion machine is impossible. A gasoline engine is not impossible. >>>If you assert that a gasoline engine can exist, and someone else accuses you of >>>trying to invent a perpetual motion machine, you can point out how the two >>>differ, and you can describe how the engine can work in practice, but if they >>>keep telling you that you are violating the laws of physics, what can you do? >>> >>>How would you like it if when I got into a sticky point with one of my >>>arguments, I wrote a big long thing declaring that you don't exist and could >>>therefore be ignored. >>> >>>Algorithm A is a single-processor algorithm designed to operate in domain D. >>>Algorithm B is a dual-processor algorithm derived from A, and applied to domain >>>D. >>> >>>There is nothing that would preclude B from being more than twice as fast as A. >>>A may not be the most optimal algorithm. Adding parallelism in order to create >>>B may have changed the algorithm significantly. The two "halves" of B may >>>interact profitably. Domain D may not be well enough understood. >>> >>>None of this results in any risk of the planet splitting open or Isaac Newton >>>being refuted. >>> >>>bruce >> >>Okay, I've said this elsewhere, but here's a good place to clear things up. >> >>Bob and I are _not_ arguing that B cannot be super-linearly faster than A. >> >>Bob and I are arguing that B's existance proves the existance of Algorithm C, a >>sequential algorithm that B will not be super-linearly faster than. If domain D >>is not well-enough understood, then it may be necessary to implement C as a >>sequential simulation of B. >> >>Bob and I are furthermore also stating that since we know that C exists and that >>we can implement C, it is not appropriate to report data comparing only A and B. >> >>Dave > >Yes, that's true, and what I'm saying is not hard to understand either. > >I am saying that the discovery of a consistent super-linear parallel result >should result in investigation, the point of which is to improve the serial >version (which will probably improve the parallel version, too). > >http://www.icdchess.com/forums/1/message.shtml?191831 > >Look at the last couple of paragraphs of mine, where I lay this out, and then >Bob's last paragraph, where he immediately disagrees with something that I am >not even saying. > >This is like being on an Escher staircase. > >bruce It certainly is, but _not_ for the reason you think. Look _carefully_ at what I wrote. First, you said "I'll bet that in a group of 1000 students one of them would..." I said "I disagree. And I gave reasons why, including the fact that I give similar things to do in my parallel programming courses here, and I have _never_ had anyone do this. They _always_ have both threads doing the same algorithm, at least for their first attempt. Because it is completely logical, and much easier to debug when you don't have two different functions, one for the first half of the list, another one for the second. So I am not "disagreeing with something you didn't say". I am saying that based on teaching a _lot_ of students over the years how to write parallel code using posix threads, PVM, MPI, HPF, and you-name-it, I have _never_ seen anyone write a parallel code like that. And not only have I _not_ seen such a piece of code written, I wouldn't expect it to be written for lots of reasons that are intuitively obvious...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.