Author: Robert Hyatt
Date: 17:48:23 09/12/99
Go up one level in this thread
On September 12, 1999 at 15:53:42, John R. Menke, Sr. wrote: >The speed (kN/s) varies considerably using the infinite analysis mode setting >with Fritz 5.32 or Hiarcs 7.32. For example, with a particular position Fritz >may calculate at a rate of 40kN/s. Then if I reset my computer and reload, or >perform various other intermediate operations, it may then calculate at >70kN/s(same position, same settings, same hash table size, etc.). That's >typical. I cannot depend upon getting the best analysis speed without trying it >several times to test it, and see what range is involved! Why? The problem is caused by direct-mapping in the cache. If your executable gets plopped into consecutive real memory pages, it gets spread evenly over the cache, and performance is best. If your executable gets put into (say) every other physical page, you only direct map to 1/2 of cache, and performance goes down. You typically get somewhere between these two extremes, although in rare cases it can be far worse than just 2x...
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.