Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: root search ordening

Author: Christophe Theron

Date: 19:45:06 05/02/01

Go up one level in this thread


On May 02, 2001 at 12:52:04, Robert Hyatt wrote:

>On May 02, 2001 at 12:40:32, Paul wrote:
>
>>On May 02, 2001 at 12:19:19, Robert Hyatt wrote:
>>
>>>On May 02, 2001 at 11:37:02, Paul wrote:
>>
>>>>I also use nodecounts after reading about it on CCC. But what type do you use
>>>>for the counter, an unsigned int? That will overflow after a search of a couple
>>>>of hours on these 1M node/s machines we have nowadays, wouldn't it? Or am I
>>>>missing something (like a shift right)?
>>>>
>>>>Groetjes,
>>>>Paul
>>>
>>>
>>>I just use an unsigned int at present.  4 billion nodes on a single root move
>>>is a bunch of searching.  Hopefully by the time it happens we will be on 64
>>>bit processors and the problem won't exist.
>>
>>Are you sure? 4 billion nodes cause overflow in just over an hour @ 1M n/s ...
>>that's pretty quick when you're analyzing one of Leonid's mates. :)
>>
>>Paul
>
>Not quite.  that is 4 billion nodes for a single move at the root.  Since each
>root move has its own node counter associated with it.  Yes it might happen on
>an overnight search, but I worry little about the odd cases, and more about
>playing in live 'games'.  ;)



Well recently I had to convert my node counter from 32 to 64 bits, because Uri
was doing overnight analysis. :)



    Christophe



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.