Author: Bernhard Bauer
Date: 07:03:54 01/20/98
Go up one level in this thread
On January 20, 1998 at 09:06:21, Robert Hyatt wrote:
>On January 20, 1998 at 02:21:09, Bernhard Bauer wrote:
>
>>On January 19, 1998 at 12:57:52, Bruce Moreland wrote:
>>
>>>
>>>On January 19, 1998 at 10:03:27, Robert Hyatt wrote:
>>>
>>>>On January 19, 1998 at 07:10:46, Bernhard Bauer wrote:
>>>
>>>>> hash depth time score trans/ref pawn used w used b nps
>>>>> 64k 31-> 34.82 2.85 79% 99% 98% b96% 109889
>>>>> 96k 31-> 23.34 3.04 53% 99% 99% b98% 125695
>>>>> 192k 31-> 25.12 3.04 66% 99% 99% b98% 122916
>>>>> 384k 31-> 17.73 3.02 74% 99% 99% b98% 126340
>>>>> 768k 31-> 9.85 3.25 90% 99% 99% b98% 125667
>>>>> 1.5M 31-> 6.50 3.25 121% 99% 79% b92% 126525
>>>>> 3M 31-> 3.17 3.25 136% 99% 27% b37% 123640
>>>>> 6M 31-> 34.66 2.85 79% 99% 98% b96% 110491
>>>>> 12M 31-> 34.25 2.85 82% 99% 88% b83% 109862
>>>>> 24M 31-> 36.02 2.85 84% 99% 68% b62% 110227
>>>
>>>1) How much memory is on the machine?
>>>
>>
>>This is a PPro/200 with 64 Mb memory running WinNT4.0.
>>
>>>2) Notice how the 6M line is almost identical with the 64K line. Isn't
>>>that suspicious?
>>
>>I tend to think so too, however, I have not enough nowledge about
>>crafty.
>>Therefor I would like to see other programs behavior on this problem.
>>May be you could show us how Ferret does?
>>
>>Kind regards
>>Bernhard
>>
>>>
>>>bruce
>
>
>something is definitely wrong. the hash usage should continue to drop
>as
>you make the hash table larger and larger. However the last 3 show that
>it
>is higher than the ones before it. I will try to run this myself in a
>bit
>and post my results on a 128mb machine...
I agree.
Here are my results for fine70 on a PP200 running WinNT4.0 64Mb.
Times to finish a ply vs hash table size for fine70 using Crafty14.4
| hash size in Kb
|
ply | 96 192 384 768 1536 3072 6144 12288
24576
|
-------------------------------------------------------------------------------
|
28 | 2.30 6.80 2.34 1.70 1.72 1.71 3.39 3.32
3.36
29 | 6.15 13.90 4.78 2.75 2.52 2.05 5.53 5.45
5.47
30 | 9.53 15.42 9.11 4.53 4.50 2.47 8.81 9.53
9.24
31 | 23.45 25.25 17.84 9.81 6.55 3.24 34.72 34.07
35.94
32 | 68.00 43.01 29.76 14.45 10.34 4.50 60.00 52.04
48.42
33 | 355.00 203.00 55.53 30.80 15.34 12.22 1006.00 873.00
701.00
34 | 158.00 90.00 34.67 21.52
35 | 144.00 73.00 41.91
36 | 1414.00 207.00
Up to 3MB everything looks ok, but then we have a jump. Hope you can
resolve
the problem, because it may be everywhere, but not as visible as here.
Kind regards
Bernhard
This page took 0.01 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.