Author: Sune Fischer
Date: 05:28:15 12/30/01
Go up one level in this thread
On December 30, 2001 at 08:06:37, Uri Blass wrote:
>>>For me makemove and generating all the attack tables cost time so I prefer to do
>>>it only after verifying that the capture is good enough.
>>
>>Why not update the attack boards _before_ generating moves, then you have a fast
>>makemove that you can do at the leaf too.
>
>The problem is that updating the attack tables is most of the time of the
>makemove so it is not going to save much.
I think we are doing almost the same actually, none of us update the attack
boards until we absolutely need to.
I do:
makemove(...)
{
// update simple information, like incremental scores and hashkeys
}
generate_moves(..)
{
//get the attack boards
//and find the legal moves...
}
Since I do not generate moves at the leaf, I don't need to find the expensive
attack boards.
Unless I need to evaluate when the position isn't hashed, then I:
evaluate(..)
{
//get the attack boards
//and evaluate....
}
>> I have no choice but to do the move
>>before evaluation or else I would simply not evaluate the right position, you
>>too must be doing some sort of move I think.
>
>The tactical change can be evaluated easily and the positional change has a
>bound so you still can see if a capture is not good enough.
>
>Uri
Yes I know, this is the very first check of my eval.
I get around 400 kNs doing this, but I have a lot of safety stuff turned on
(like memsetting), so I am optimistic.
-S.
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.