Comments on: “Jnuchess” Plays and Wins http://www.forthgo.com/blog/2005/05/23/jnuchess-plays-and-wins/ Code and Recreations of Xan Gregg Sat, 17 Jul 2010 15:38:11 -0700 http://wordpress.org/?v=2.9.2 hourly 1 By: xan http://www.forthgo.com/blog/2005/05/23/jnuchess-plays-and-wins/comment-page-1/#comment-10 xan Tue, 24 May 2005 15:25:25 +0000 http://www.forthgo.com/blog/2005/05/22/jnuchess-plays-and-wins/#comment-10 Thanks Anli. I'm still just using a transliteration of the C code, which manages the hash table "manually". When I get to the point to Java-izing the code more, I'll check out the Trove work. The hash table maps board positions to evaluation scores, and there are several interesting problems besides the mechanics of the table. For example, how to most efficiently represent a chess position, and whether some positions are more cache-worthy than others. But first I need to get the straight port done... Thanks Anli. I’m still just using a transliteration of the C code, which manages the hash table “manually”. When I get to the point to Java-izing the code more, I’ll check out the Trove work.

The hash table maps board positions to evaluation scores, and there are several interesting problems besides the mechanics of the table. For example, how to most efficiently represent a chess position, and whether some positions are more cache-worthy than others.

But first I need to get the straight port done…

]]> By: Anli http://www.forthgo.com/blog/2005/05/23/jnuchess-plays-and-wins/comment-page-1/#comment-9 Anli Tue, 24 May 2005 13:23:26 +0000 http://www.forthgo.com/blog/2005/05/22/jnuchess-plays-and-wins/#comment-9 I'm curious about the hash table management problem. Are you using 1.4's HashMap-s? They rehash the returned hashCode-s. I don't know whether that's overkill if your hashCodes are already well distributed. Another alternative might be to use Trove's hashMaps: he uses another hashing algorithm (Knuth's prime-number sized arrays with overlapping lookups) and also maps of ints etc. It's LGPL code so the execs don't want to hear about it... but you don't have that constraint. I’m curious about the hash table management problem. Are you using 1.4’s HashMap-s? They rehash the returned hashCode-s. I don’t know whether that’s overkill if your hashCodes are already well distributed.

Another alternative might be to use Trove’s hashMaps: he uses another hashing algorithm (Knuth’s prime-number sized arrays with overlapping lookups) and also maps of ints etc. It’s LGPL code so the execs don’t want to hear about it… but you don’t have that constraint.

]]>