Author: Ren Wu
Date: 09:09:25 10/27/99
Go up one level in this thread
On October 25, 1999 at 23:25:11, Eugene Nalimov wrote: >Nothing unusual by today standards... Dual CPU 500Mhz computer. The only "hard" >requirement is >=1Gb RAM. And of course dosk space to save uncompressed TB after >it's generated. > >Time to generate one TB depend on a lot of factors: TB size (KxyKzz is 4 times >larger than KxxKxx; 8 times, if it's 16-bit table), percentage of draws (the >higher is it, the more work had to be done on each iteration), # of iterations >(now it's equal max DTM+1). So, all KxxKxx were generated in 1-2 days. Average >KxxKyz needs a week or so (my estimate). "Record" TB (KRNKNN, KRBKNN) will need >~2 weeks. And of course now I am trying to generate "record" TBs. > >Currently generator will run only if compiled by 32-bit compiler, as I had no >time to clean it up to make it independent of sizeof (unsigned int). It's not >possible to generate large TBs under 32-bit Unix (e.g. Linux) because of the 2Gb >file limitation those dumb operating systems have. And we had no time to fix >DATACOMP, so it still cannot handle very large files. So, for now maximum that >can be done with current code base is pawnless KxxKyz, and largest TBs have to >be produced on the machine running "real" OS - i.e. Windows NT :-). > Or BeOS! :) This is very good, thanks very much Eudene. Ren.
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.