Author: Jeremiah Penery
Date: 18:10:07 06/01/99
Go up one level in this thread
On June 01, 1999 at 19:47:53, Dann Corbit wrote: >On June 01, 1999 at 19:39:28, Dave Gomboc wrote: > >>On June 01, 1999 at 19:00:28, Dann Corbit wrote: >> >>>Larry Applebaum's settings and build file seem to have created the fastest >>>version. I have changed my file concatenation order to be the same as his, and >>>now I get the same optimal performance as his wcraftyz.exe. You can find it on >>>my website as: >>> >>>ftp://38.168.214.175/pub/Crafty1.exe >>> >>>If you got a copy over the weekend, get a new one. This one is 6% faster than >>>my previous version, as I now get 170,000 NPS on a PII 300 and 281,000 NPS on a >>>PIII 500. >> >>This is a MAJOR speed improvement! A crafty SMP version (but with mt=1) of 16.6 >>that I have been using gets about 100000 NPS on a PII 300 (NT). >To get 170K, you need to do a start/high crafty1.exe >I don't recommend start/realtime, since it makes the machine unusable for >everything else, and can even be hard to get control back. Besides which, it >does not seem to increase the bench numbers any more over start/high. >My SMP version is slightly slower, but should still be a major improvement over >what you are getting now. If you only run one CPU, it is better to use the >non-smp version, since the MT libraries are not as fast as single threaded. I always got around 170k NPS on a PII-300 using NT, ever since version 16.5 (with the inline assembly). As long as you're not running anything else, even start/low should get nearly the same numbers. Jeremiah
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.