Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty and NUMA

Author: Robert Hyatt

Date: 07:44:37 09/04/03

Go up one level in this thread


On September 04, 2003 at 03:56:07, Uri Blass wrote:

>On September 04, 2003 at 03:44:22, Johan de Koning wrote:
>
>>On September 03, 2003 at 09:08:46, Vincent Diepeveen wrote:
>>
>>[to recurse or not to recurse]
>>
>>>Let's talk about the good aspects, i save out a lot of expensive function calls.
>>
>>Funny. :-)
>>
>>>If i remember well The King isn't recursive either. Johan no doubt reads this
>>>and will say: "no way" when it isn't :)
>>
>>No Way!
>>I'm not sure if I just confirmed your memory or if I denied it.
>>
>>Now please stop calling my name, or Frans', or whoever's, in threads that we
>>don't want to read at all. Since "we know shit" about parallelorisming we won't
>>be able to back you up. Try evaluation or TT latency. :-)
>
>I think that the subject of being recursive or not being recursive
>is important not only for parallelorism.
>
>For me it is more interesting if there are programmers who tried recursive and
>non recursive for one processor and if they got speed improvement from the
>non recursive possibility and how much speed improvement did they get from it.
>
>Uri


I did both on the Cray.  I couldn't tell any difference in terms of speed.

A procedure call is _not_ slow on all machines.  When I started Crafty, I
first wrote the search and stuck it into Cray Blitz to make sure I understood
the negamax stuff correctly.  When I noticed that the speed didn't change, I
was much happier with the negamax approach for simplicity, and I kept it.  I
would prefer the iterated approach for parallel search, and one day I might do
it again, but for the moment, what I have works just fine.



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.