Author: Larry Griffiths
Date: 19:06:12 08/15/00
Go up one level in this thread
On August 15, 2000 at 21:03:30, Tom Kerrigan wrote:
>I just ran this experiment with my program.
>
>Ordering killers before captures was 20% slower than "normal," in terms of total
>nodes searched.
>
>-Tom
Tom,
I ran my program with and without killers. The board is at initial position and
my killer code is somewhat broken... :)
WITHOUT KILLERS...
Elapsed Time= 136.08 seconds
Ply ---CAPTURES-- -----MOVES----- -----TOTAL----- BF ---Invalid---
1 0 80 80 13 0
2 0 1,112 1,112 8 0
3 63 9,043 9,106 13 0
4 1,224 119,468 120,692 4 50
5 17,086 504,270 521,356 11 2,756
6 210,422 5,909,744 6,120,166 3 13,128
7 2,119,318 17,561,799 19,681,117 6 659,596
8 15,388,259 121,767,397 137,155,656 0 1,547,087
_____________ _______________ _______________ _____________
17,736,372 145,872,913 163,609,285 2,222,617
Captures/Moves Per Second= 1,202,320
WITH KILLERS...
Elapsed Time= 102.19 seconds
Ply ---CAPTURES-- -----MOVES----- -----TOTAL----- BF ---Invalid---
1 0 80 80 13 0
2 0 1,078 1,078 7 0
3 62 7,840 7,902 13 0
4 1,300 102,139 103,439 3 66
5 13,936 361,277 375,213 11 3,707
6 172,004 4,194,997 4,367,001 3 11,852
7 1,684,612 11,819,330 13,503,942 7 695,601
8 10,618,615 90,809,707 101,428,322 0 1,358,624
_____________ _______________ _______________ _____________
12,490,529 107,296,448 119,786,977 2,069,850
Captures/Moves Per Second= 1,172,233
I think I got even better results with a middlegame position, but I will have to
run a test and see what happens. I will also have to work on putting killers
before captures and moves and see what happens.
Larry.
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.