Author: Robert Hyatt
Date: 10:45:44 10/05/99
Go up one level in this thread
On October 05, 1999 at 12:28:46, Bruce Moreland wrote:
>
>On October 05, 1999 at 05:25:21, Bas Hamstra wrote:
>
>>What are good ways to cut down the number of evals? I saw Bob Hyatt post that he
>>could easily double NPS when using "Lazy Eval".
>>
>>What is a correct way to do that? Is there more to it than the qsearch "delta"
>>type of pruning?
>
>He couldn't double NPS, that's crazy.
>
>bruce
No... and I didn't read what he wrote there carefully when I responded to
it. I generally do a full, slow eval on about 10% of the positions I search,
which means that the lazy eval cuts out about 90% of the time before doing the
whole thing.. and even so, my eval is taking 40-50% of the total search time
when profiled. I don't know what turning lazy eval off will do, but here is a
quick run:
kopec22 depth=11 23.19 secs with lazy eval nps=172K
32.15 secs without lazy eval nps=123K
kopec3 depth=11 106 secs with lazy eval nps=152K
135 secs without lazy eval nps=119K
So it seems to be worth something around 50%... ie I go 50% slower without
it than I do with...
on two test positions, of course...
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.