Author: Robert Hyatt
Date: 11:04:49 12/12/02
Go up one level in this thread
On December 12, 2002 at 10:15:16, Vincent Diepeveen wrote: >On December 11, 2002 at 02:34:33, Matt Taylor wrote: > >[snip] >>Eugene's explanation fits, though. I am suprised that Intel did not duplicate >>the trace cache for both logical CPUs. It's like trying to fit an even bigger >>peg into an already too small hole... >>-Matt > >Exactly, but the hardware reason to do that is very simply. > >They can clock the thing to 3.04Ghz now. 2.8Ghz for the Xeon. > >But if you double the L1 data cache size or the trace cache size >(i will not do a statement what in my eyes is smarter to duplicate >because you can see my next sentence why) then you have a major other >problem. > >You won't be able to clock it to 3.04Ghz then nor 2.8Ghz for the Xeon. > >If you have something small, you can clock it high. > >If you have something big like an Itanium2 or the 128KB L1 cache >of a K7 then you can't clock it that easily to 3.04Ghz. > >So the clocking and the size of such important integrated things into >the procesor is very closely related. > >Of course i didn't checkout the P4 manual too well yet >(it's so big and there are so many manuals), apart from what >i already posted about it; but i still wonder how they plan >to replace a spinlock in crafty. I find it a theoretic discussion. > >Of course Bob can start using the Huber (abdada or whatever) algorithm >to parallellize his engine (no locking needed) but i know Bob also posted >very negative things about that being no real parallel algorithm. > >So i wonder how Bob is going to rid of locking within crafty without >a year or 2 of rewriting crafty by some professional parallel programmer. First, let's get a few points out of the way. I _am_ a professional parallel programmer. I have written at _least_ 100X as much parallel code as you have, and, in fact, I spent most of a year explaining parallel search to you. I was doing parallel programming in the 1970's. I had a parallel search in 1978 at the ACM event, running on a Univac 1100/42 dual-cpu machine. Second, I am _not_ trying to eliminate spinlocks. They are necessary and they can _not_ be eliminated. I won't explain why, but any book on parallel programming will give you all the info you need to understand why. Third, spinlocks are not bad. The smp_lock in crafty is not bad. Regardless of how many times you say it is. The reason it is not bad is that it is not used very much. If I split 2000 times over 3 minutes, I use that smp_lock 2000 times during that 3 minute period. That is not even measurable as time used. it is irrelevant. It has been irrelevant. And it will be irrelevant until I run on a machine with a large (large == 32 or more) number of processors... Any time you want to compare programming skills, I'm ready. And it won't be much of a contest, IMHO. Just listening to your rambling makes that point pretty obvious to most anybody. Please find another way to discuss things without (a) citing nonsensical data as fact; (b) referring to other programs as always inferior to yours; (c) referring to yourself as the best programmer around. None of those are true, and stating them over and over will _not_ change that... > >Best regards, >Vincent
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.