Author: martin fierz
Date: 06:06:14 04/07/04
Go up one level in this thread
On April 07, 2004 at 08:33:59, Renze Steenhuisen wrote: >On April 07, 2004 at 08:27:59, martin fierz wrote: > >>On April 07, 2004 at 06:49:59, Renze Steenhuisen wrote: >> >>> >>>Hi all, >>> >>>could someone give me some numbers that are common with hashkey collisions? >>>Because I guess my % is little too high... >>> >>>I'm getting like 0.03% [which is 1 every 3000, if I'm not mistaken] >>> >>>This is when using TT=32MB (haven't got the exact number of entries) >>> >>>If you think it is an error, any suggestions on where to start looking? >>> >>>Thanks! >>> >>> Renze >> >>hi renze, >> >>your number is much too high. a good estimate for the hash collision probability >>is 1/sqrt(hashkey_range); which in your case comes out as 2^-32 or about >>somewhere around one in a billion... >> >>make sure you're not doing anything wrong with the hashkey updating like >>suggested by others - compare with a computed-from-scratch key. also make sure >>that your nullmove doesn't break your hashkey. and just to be sure, check that >>you haven't got 32-bit numbers instead of 64 bit numbers by accident... >> >>cheers >> martin > >I am using uint64_t for that (standard C), so it is 64-bit... just make sure :-) i once thought i had 32-bit keys in an old program of mine until i discovered that rand() returns 16-bit values... whenever i generate hash keys now, i just do a quick dump of them to a file in binary notation to verify that they look good, i.e. look sort of random and that all bits are being used. if there is something wrong with the numbers you see it immediately. cheers martin > >I'm going to verify (just to make sure) the incremental with the from-scratch >key, so I'll post with in the hour an update... > >one in a billion was what I recollected from previous posts (long time ago), so >I remembered correctly > >Thanks! > Renze
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.