Author: Vincent Diepeveen
Date: 05:37:09 06/06/01
Recursive double nullmove + 150mb hash + 8 probes in hashtable + good qsearch.
Material only, but in my material count a bishop is worth more as knight,
so it hunges basically with knight after bishops! And bishop captures knight
it always tries, though this is now stupid to try in qsearch.
Not a single extension is getting done except checks in qsearch are
getting done until mate in 16 above mainsearch depth or something.
Note below is a debug version at P3-800 run dual.
Proces 1 starting initialisation!
Proces 1 starting SearchSMP!
Took 0.07 seconds to start all 1 other processes
Process 1: engineflags=0 msk=1
process 0: engineflags = 0 denktime=10000000 maxtime=10000000
00:00 0 0 3 (0) 1 0.000 Ng1-f3
00:00 0 0 44 (0) 2 0.000 Ng1-f3 Ng8-h6
00:00 0 0 145 (0) 3 0.000 Ng1-f3 Ng8-h6 Rh1-g1
00:00 0 0 287 (0) 4 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
00:00 0 0 471 (0) 5 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
00:00 0 0 1582 (0) 6 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8
00:00 0 0 2384 (0) 7 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8 Rh1-g1
00:00 0 0 4251 (0) 8 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8 Rh1-g1
Rh8-g8
00:00 0 0 6916 (0) 9 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8 Rh1-g1
Rh8-g8
00:00 0 0 10386 (0) 10 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8 Rh1-g1
Rh8-g8
00:00 0 0 30058 (0) 11 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8 Rh1-g1
Rh8-g8
00:00 0 0 44288 (0) 12 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8 Rh1-g1
Rh8-g8
00:00 0 0 81591 (307) 13 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8
Rh1-g1Rh8-g8
00:00 0 0 149159 (5312) 14 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8
Rh1-g1 Rh8-g8
00:02 0 0 412505 (7476) 15 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8
Rh1-g1 Rh8-g8
00:05 0 0 1002334 (12353) 16 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8
Rh1-g1 Rh8-g8
00:09 0 0 2077733 (12851) 17 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8
Rh1-g1 Rh8-g8
00:21 0 0 4487766 (371847) 18 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8
Rh1-g1 Rh8-g8
01:08 0 0 13712688 (1973228) 19 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Rg8-h8Rh1-g1 Rh8-g8
03:04 0 0 36269651 (3996136) 20 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Rg8-h8Rh1-g1 Rh8-g8
09:53 0 0 111944574 (28593783) 21 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Rg8-h8 Rh1-g1 Rh8-g8
14:08 0 0 162171653 (28666944) 22 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Rg8-h8 Rh1-g1 Rh8-g8
26:25 0 0 297586092 (57442017) 23 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Rg8-h8 Rh1-g1 Rh8-g8
01:17:09 0 0 765321662 (142518442) 24 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Rg8-h8 Rh1-g1 Rh8-g8
02:37:45 0 0 1591024593 (165276208) 25 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
Rg1-h1Rg8-h8 Rh1-g1 Rh8-g8
Now exactly the same debug version without nullmove, without extensions, so
completely FULLWIDTH.
NO NULLMOVES!
here is fullwidth output:
Took 0.08 seconds to start all 1 other processes
Process 1: engineflags=1 msk=1
process 0: engineflags = 1 denktime=10000000 maxtime=10000000
00:00 0 0 3 (0) 1 0.000 Ng1-f3
00:00 0 0 44 (0) 2 0.000 Ng1-f3 Ng8-h6
00:00 0 0 145 (0) 3 0.000 Ng1-f3 Ng8-h6 Rh1-g1
00:00 0 0 1136 (0) 4 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
00:00 0 0 2881 (0) 5 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
00:00 0 0 18254 (0) 6 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
00:00 0 0 44132 (0) 7 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6 Rh1-g1
00:01 0 0 263338 (195) 8 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6 Rh1-g1
Nc6-b8
00:04 0 0 656751 (218) 9 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6 Rh1-g1
Nc6-b8
00:14 0 0 3501150 (218) 10 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
Rh1-g1 Nc6-b8
00:48 0 0 8852828 (218) 11 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
Rh1-g1 Nc6-b8
03:17 0 0 44895497 (218) 12 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
Rh1-g1 Nc6-b8
11:41 0 0 121046950 (218) 13 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
Rh1-g1 Nc6-b8
52:34 0 0 668413587 (577) 14 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
Rh1-g1 Nc6-b8
03:29:01 0 0 2117917129 (2236) 15 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Nb8-c6 Rh1-g1 Nc6-b8
i stopped 16 ply search. it was still busy with PV and already searching for 15
hours when
i stopped it. As i print out every second what it's busy with i have good
reasons to assume
that it would have required another couple of hours up to 10 hours perhaps to
get the PV for
16 ply.
Here the log output:
00:00 0 0 3 (0) 1 0.000 Ng1-f3
1 22 0.01
00:00 0 0 44 (0) 2 0.000 Ng1-f3 Ng8-h6
2 82 0.01
00:00 0 0 145 (0) 3 0.000 Ng1-f3 Ng8-h6 Rh1-g1
3 608 0.03
00:00 0 0 1136 (0) 4 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
4 1674 0.05
00:00 0 0 2881 (0) 5 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
5 9835 0.11
00:00 0 0 18254 (0) 6 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
6 25578 0.25
00:00 0 0 44132 (0) 7 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
Rh1-g1
7 139118 0.86
00:01 0 0 263338 (195) 8 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
Rh1-g1 Nc6-b8
8 375557 2.50
00:04 0 0 656751 (218) 9 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Nb8-c6
Rh1-g1 Nc6-b8
9 1956141 8.45
00:14 0 0 3501150 (218) 10 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Nb8-c6 Rh1-g1 Nc6-b8
10 5045672 27.76
00:48 0 0 8852828 (218) 11 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Nb8-c6 Rh1-g1 Nc6-b8
11 25483059 108.75
03:17 0 0 44895497 (218) 12 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Nb8-c6 Rh1-g1 Nc6-b8
12 67548506 391.69
11:41 0 0 121046950 (218) 13 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Nb8-c6 Rh1-g1 Nc6-b8
13 367000147 1698.14
52:34 0 0 668413587 (577) 14 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Nb8-c6 Rh1-g1 Nc6-b8
14 1098209967 6655.11
03:29:01 0 0 2117917129 (2236) 15 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
Rg1-h1 Nb8-c6 Rh1-g1 Nc6-b8
15 7096237081 33227.48
Now fullwidth no extensions, last 6 ply no hashtable. so no store and no
retreive either.
Proces 1 starting SearchSMP!
Took 0.06 seconds to start all 1 other processes
Process 1: engineflags=1 msk=1
process 0: engineflags = 1 denktime=10000000 maxtime=10000000
00:00 0 0 3 (0) 1 0.000 Ng1-f3
00:00 0 0 44 (0) 2 0.000 Ng1-f3
00:00 0 0 145 (0) 3 0.000 Ng1-f3
00:00 0 0 1136 (0) 4 0.000 Ng1-f3
00:00 0 0 3517 (0) 5 0.000 Ng1-f3
00:00 0 0 28462 (0) 6 0.000 Ng1-f3
00:00 0 0 89189 (0) 7 0.000 Ng1-f3
00:02 0 0 775848 (0) 8 0.000 Ng1-f3 Ng8-h6
00:11 0 0 2482606 (113) 9 0.000 Ng1-f3 Ng8-h6 Rh1-g1
01:03 0 0 22362772 (943) 10 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
04:26 0 0 60875158 (66610) 11 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
22:27 0 0 443945289 (68599) 12 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-h8
01:25:42 0 0 1170110639 (70341) 13 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8-
h8 Rh1-g1
06:53:42 0 0 7734566503 (105403) 14 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1 Rg8
-h8 Rh1-g1 Rh8-g8
busy searching mainline of 15 ply. Searched for 24 hours. Stopped it now. It
still was
busy with Nf3 Nh6 so it would take a long time before all the second moves after
the root
would be finished searching. Each tree nearly as big of course as the tree it
had
searched. So estimation would be 50% x 19 moves x another (24-7) hours = 9.5 x
17 hours.
For sure many days.
Here the log file output of th efullwidth search. it needed 12 billion nodes
to finish 14 ply:
00:00 0 0 3 (0) 1 0.000 Ng1-f3
1 22 0.04
00:00 0 0 44 (0) 2 0.000 Ng1-f3
2 82 0.04
00:00 0 0 145 (0) 3 0.000 Ng1-f3
3 608 0.05
00:00 0 0 1136 (0) 4 0.000 Ng1-f3
4 2027 0.07
00:00 0 0 3517 (0) 5 0.000 Ng1-f3
5 15223 0.11
00:00 0 0 28462 (0) 6 0.000 Ng1-f3
6 51044 0.30
00:00 0 0 89189 (0) 7 0.000 Ng1-f3
7 411532 1.24
00:02 0 0 775848 (0) 8 0.000 Ng1-f3 Ng8-h6
8 1419591 6.38
00:11 0 0 2482606 (113) 9 0.000 Ng1-f3 Ng8-h6 Rh1-g1
9 11898175 34.65
01:03 0 0 22362772 (943) 10 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
10 34287606 140.19
04:26 0 0 60875158 (66610) 11 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
11 233207916 718.55
22:27 0 0 443945289 (68599) 12 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8 Rg1-h1
Rg8-h8
12 672356151 2825.71
01:25:42 0 0 1170110639 (70341) 13 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
Rg1-h1 Rg8-h8 Rh1-g1
13 4175277805 13614.69
06:53:42 0 0 7734566503 (105403) 14 0.000 Ng1-f3 Ng8-h6 Rh1-g1 Rh8-g8
Rg1-h1 Rg8-h8 Rh1-g1 Rh8-g8
14 11993687680 52187.21
So to finish from the openings position with the most simplistic evaluation
there
is (material only, of course always returning a 0 is even simpler but
then you don't need a qsearch anymore) a 14 ply fullwidth search, without
*any* extension except a few things in qsearch get done. That's it.
Last 6 plies DIEP didn't use a transpositoin table. See how many more nps i get
then when getting to those speeds of hundreds of k's a second the hashtable
is really a bottleneck!
But compare the output to the output needed with the number of nodes needed
*with* hashtable.
The difference is HUGE!
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.