Author: Vincent Diepeveen
Date: 08:24:24 03/06/03
Go up one level in this thread
On March 06, 2003 at 11:14:43, Robert Hyatt wrote: >On March 06, 2003 at 09:34:28, Vincent Diepeveen wrote: > >>On March 06, 2003 at 09:16:22, Vincent Diepeveen wrote: >> >>I been searching in crafty source code for Lock() and see now that the inline >>assembly that used to be there for it, is REMOVED. This is shocking! >> >>It has been replaced by slow system functions instead which simply make the >>thing slower. This is what i call the usual scientific idea of making a program >>slower in order to get a better speedup in this case for SMT/HT? > >Did you notice the keyword "inline"? :) > >Didn't think so. You are trying to act like a smart politician now. wrong try. There is nothing to misunderstand here. With crafty you are doing slowly the same thing from which you accused a few years ago the Zugzwang team. > >> >>Especially with coming Opteron NUMA systems in mind it is of course a very bad >>idea to do this. >> >>Nothing as weak as system functions, and crafty *does* do a big number of >>searches each second each cpu, so locking is a very important thing to not slow >>down! > >Correct, which is why I do it as I do it. > > >> >>>the windows makefile doesn't work by default for compilers >>>when you turn on: >>> >>>COPTS = /DSMP /DCPUS=2 >>> >>>the problem of that is that then in egtb.cpp the function Lock is not defined. >>>Of course i know how to solve it, but it should be in egtb.cpp simply.
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.