Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Programming question,some solutions

Author: James Robertson

Date: 07:54:45 03/20/99

Go up one level in this thread


On March 20, 1999 at 09:27:18, Inmann Werner wrote:

>Hello.
>
>After sleeping about it and going again on to the problem, I found out, that
>most of you are right and want to thank for your suggestions!!!
>
>1) Making the new code, I started an old bug to work.
>   In Hash adressing, it is important, that hash size is something like
>   2^x (65536,131072,262144....) I did exactly this but with one error.
>   To make the Hash you have to do something like
>   Hashadress=hashsize & hashindex. And here, hashsize has to be one less than
>   above (65535,131071....). the "-1" dropped out because of the new code.
>Before, I had a special define for it, what now was obsolete.
>
>   After this bug,the lack only is about 3 to 5 %.
>   I do not understand, why it worked better on my Cyrix. Maybe another "test
>bug"?
>
>   Please excuse me, to have worried you!!!!
>
>2) Robert is right.
>   Using structures increases again, but not much. Now I only have a lack
>   of 1% to 2%.
>   I think, I can live with it, cause on my Cyrix, the code "new" now is
>   faster than the old one. This must have to do with processor
>architecture..... I am really no profi in this.
>
>Excuse again. Yesterday I really was worried. I checked the code often, what I
>changed. But I did not think, that an old bug will get started, because of the
>new thing.
>Hashing really is a "hot" and very "fragile" thing!
>
>Werner

I have had no end of troubles trying to implement hash tables. It is really
wierd because the pawn hash I finished in a few hours, while the main hash has
gotten exactly nowhere after days of work.....

James



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.