Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Books

Author: James Robertson

Date: 17:34:14 04/20/99

Go up one level in this thread


On April 20, 1999 at 19:44:35, Peter McKenzie wrote:

>On April 20, 1999 at 19:21:07, James Robertson wrote:
>
>>On April 20, 1999 at 11:00:08, James Robertson wrote:
>>
>>>I have become dissatisfied with the way my program searches for moves in it's
>>>book; it simply takes too long. How do other programs do this? If I look at the
>>>Crafty bookmaking code, I see stuff about "clusters". What does this mean? How
>>>do EXchess, or Comet or other programs search their books?
>>>
>>>Thanks for any help!
>>>James
>>
>>Thanks guys for your responses; let me think about them for a bit. I forgot to
>>mention: my book is a list of hash codes that are sorted numerically. My engine
>>jumps around through the numbers until it zeroes in on the hash code that
>>matches. If there is a match, then that move is in the book. This is wonderful
>>for small books but gets bad fast with larger ones. For instance, my program
>>often can only make three book moves a second, and it's book is just 413k (26418
>>entries).
>
>A binary search in memory should let you do this much much faster then 3 every
>second - perhaps its not the book that is slowing you down?  Or are you doing
>the binary search on a file?

I am doing the search on a file.

>Even then, with suitable caching, you should get
>much better speed than reported.

I don't know how to do anything to the caching, so whatever the iostream does by
default, that is what is happening.

>How are you timing this - using ICC?  If so, maybe there are other bottlenecks
>involved...

No... this is straight from the console....

James

>
>Peter
>
>>
>>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.