Author: Robert Hyatt
Date: 08:52:53 05/07/04
Go up one level in this thread
On May 07, 2004 at 04:30:09, Vincent Diepeveen wrote: >On May 06, 2004 at 12:45:43, Robert Hyatt wrote: > >Bob let's get realistic. > >BK is a flawed testset to test parallel speedups at. No it isn't. In fact it is a reasonable one because so many others have used it and it provides a common ground for comparing speedup. > >What happened is. You took 4 positions from that to 'proof' your 3.1. No I didn't. I ran them _all_. (The CB positions). When you first broached the subject that null is much worse than non-null, I did a quick 4 position run and found no significant difference. You then said "Bob never even ran the test" and I quickly supplied the short log file to show otherwise. You didn't like that test set and said "run the DTS (CB) positions" It will show that Crafty only gets a speedup of 1.1 on a dual. I ran them and got a normal speedup and there the argument went... I don't use 4 positions to prove any point other than you were wrong (as usual) when you said I had not even _tested_ the null vs non-null hypothesis... > >That gets disproven then by GCP doing statistical math where you know really >less than a waterbird from. You do not even realize what the +- behind every >measured speedup means. In fact you never provide them yourself. No I don't. But I _have_ given lots of data that _shows_ the variability. > >GCP then is doing different tests at your quads and very clearly determines 2.8 >using a-symmetric king safety. > There goes your ignorance again. "clearly determines 2.8" is simply wrong. It varies each and every time the test is run. I gave Martin 4 four-processor runs. Each speedup was different. Yet you continue with the "2.8 is the right and only right number." Talk about "pig-headed"... All of the positions I ran were in normal game mode. All the asymmetry was on. Everything was just like it is when playing in a real game... So please grow up and stop the "clearly determines 2.8". My log that I sent you "clearly determined 3.0 and 3.1". >My 1.0 out of 2 speedup for crafty came of course from using tests with >symmetric king safety. And I (nor anyone else) has reproduced that. I run in analysis mode all the time. Give me a position or N positions. I'll do the same test. Your 1.0 was just garbage... > >But GCP using 30 positions gets down to 2.8 speedup. > >And he gets one time 3.2 and another time 3.0 for speedup when searching >fullwidth. > >a) this shows how poor your way of representing things is. You are just showing >things like you want them to see Are you talking about me or about you? I post real data. You _never_ posted any data showing your 1.0 speedup. You never post any data of any kind. Just tons of garbage. I _do_ provide real data... > >b) i do not believe a crap you did a honest test where at position 21 in BK the >crafty on average gets a 2.4 speedup when searching fullwidth. No matter how i >test fullwidth at position 21, speedups are better there for fullwidth using >asymmetric king safety. That makes us even. I don't believe a thing you say either. Fortunately, I provide data to support what I say. You provide hand-waving... > >You did not do a honest testing to get to your numbers, even in those 4 >positions and yet you deny GCP's testing at 30 positions which were honestly >done, to be true. I didn't deny GCP's testing. I'll challenge you again, just as I did about the JICCA article you keep forgetting to cite. Show me the post where I _ever_ said his testing on my quad was invalid. I simply said the results are variable. Martin found a variance of +/- .2 on the first 24 DTS positions in the logs I made available. Somehow you can't seem to grasp the fact that there _is_ real variability... But that is _your_ problem, not mine... > >This shows your true nature. > I hope it does. Of course it also shows _yours_. Where is the JICCA article? Where is the quote where I said my speedup formula works beyond 8 processors? All lies. All Fraud. All done by _you_. And you talk about "my true nature"??? Try a mirror on occasion...
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.