Author: Daniel Clausen
Date: 21:36:26 12/31/99
Go up one level in this thread
Hi On December 31, 1999 at 18:38:49, Landon Rabern wrote: >I just changed my search to remove checks on the next recursive level if the >king was captured instead of running a check procedure inside the search. How >do I remove checks at the leaves if I don't generate captures there? Well.. you *have* to do something in the leaves. :) Either normal check-test at the beginning of the search or make sure you create all captures in the qsearch. I did this 'king-capture' trick too in my old engine to avoid the CheckTest() calls, but since I still had to do this call in the leaves it didn't really save any time... >one I did last year in pascal for my high school senior programming project. I >have not implemented the q-search yet, will this fix the problem? Since the >q-search is already imperfect it won't really matter much if I get that problem >at the last depth of it, will it? Q-search may be imperfect, but then everything except an exhaustive search is. I'm pretty sure you'll get quite a lot of useless results if you accept illegal positions as leaves. >I am getting around 180,000 nodes per second. Is that slow of fast or somewhere >in between? Now this is surely hardware-dependent. What HW do you use? Is this with or without node-evaluation? What kind of position is it? (Nodecounts tend to grow in the endgame for example) And keep in mind, that NPS (nodes per second) is just as good (prolly worse) as MHz for comparing processor speed. Ie you don't get anywhere if your branching factor is twice as big than another engine. Also, not all programmers count NPS the same way. Ie you can include inner nodes also or only the leaves. Or you count these illegal nodes too or you don't. I'm also working on an engine, and for the moment I'm simply not interested in any constant speedup-factor. Kind regards, -sargon PS. [Can't believe I post to CCC at 6:35 local time... am I *that* drunk?]
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.