Author: Robert Hyatt
Date: 17:06:41 12/10/02
Go up one level in this thread
On December 10, 2002 at 19:45:44, Vincent Diepeveen wrote: >On December 10, 2002 at 16:43:29, Matt Taylor wrote: > >bob is claiming the wrong thing towards you. > >chance a process is spinning is very small with 2 processors. What on earth are you talking about? When I start a search with four threads, _three_ are spinning like mad until the fourth decides to split the tree and give the others some work. This happens _every_ time a processor finishes its work at a split point. It returns to ThreadWait() and sits in a spin-wait that is basically while (!something_to_do); > >if not then at a dual machine bob's thing would get a very bad >speedup like 0.9 at 2 cpu's or so instead of the claimed 1.7 >speedup (though my own data shows 1.6 and bob's data is not >accurate as he says himself because he only ran the test once >at his machine so every datapoint can be far off; just like >that his 5 positions that show here that he gets a positive >speedup, without doing a counter example, is proving a speedup >and 6 tests that proof the opposite is not accurate enough >info). That is about the most garbled logic I have ever seen... But it doesn't matter. Crafty has spinwait in the idle threads. Whether you like it or not... And it does _nothing_ to harm performance, so long as the machine is used only for Crafty, as I have explained several dozen times, to those that take the time to listen... > >Still understanding what i write? > >No it's just Hyatt who tries to ignore the truth here. Yes, I ignore _your_ truth. Which is not the real-world truth. > >truth is that in this world the only big speedups with SMT >is reported by wintel and intel guys. I reported my number. Eugene reported his number. Others will eventually report _their_ numbers. And HT _does_ work... > >Bob is one of them, that's my claim. > >Bob is using machines from nalimov which are in the wintel >labs, to proof SMT works. I am using a machine sitting right in my office, so once again I have absolutely no idea what you are talking about or why you are talking about it. Do _you_ have a hyper-threading machine? I do. > >This where the intel documentation clearly shows that if you >buy in a shop a 2.8Ghz Xeon, that it doesn't have working SMT >at all. > That is utter baloney. Want to see the output from my linux kernel showing the dual Xeon 2.8 has four logical processors? I'd be happy to post it to once again _prove_ what you write is just wild rambling nonsense... > >I quoted a week ago here the exact reference to the PDF of the >intel architecture guide. So? I don't need the _reference_. I have the _machine_. A Dell PowerEdge 2600 with dual 2.8's... sitting right in my office at this second...
This page took 0.01 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.