Author: Christophe Theron
Date: 09:22:42 02/20/02
Go up one level in this thread
On February 20, 2002 at 08:11:30, Gian-Carlo Pascutto wrote:
>I'm wondering if Chess Tiger could have problems when
>being used for overnight analysis.
>
>I put up the Deep Blue - Kasparov position in ChessParter 5.1,
>started ChessTiger with 200M hashtables (the maximum) on the
>XP 1800 and let it analyse away.
>
>However, after about 1.5 hour the nodecounter overflowed,
>and the next PV that was printed out was cut short, and
>the branching factor exploded. Either it might be on to
>something, or it might be having problems with overloaded
>hashtables.
>
>Anybody else seen this kind of problem in very long analysis runs?
>
>00:00:00.0 -1,14 11 164841 Qxc6 dxc6 Kg8 Ra7 Rc8 Rb7 Ra8 Rxb5 Ra1+ Kf2 Ra2+ Kg1
>Ra1+ Kh2
>00:00:00.4 -1,58 11 193010 Qxc6 dxc6 Kg8 Ra7 Rc8 Rb7 Kf8 Rxb5 Ra8 Rd5 Ra1+ Kf2
>Ra2+ Kg1
>00:00:00.5 -1,58 12 389392 Qxc6 dxc6 Kg8 Ra7 Rc8 Rb7 Kf8 Rxb5 Ra8 Rd5 Ra1+ Kf2
>Ra2+ Kg1 Bc7
>00:00:03.2 -2,34 13 1412753 Qxc6 dxc6 Rc8 Ra7+ Kf8 Bd5 Rc7 Ra8+ Ke7 Ra5 Ke8
>00:00:09.3 -2,34 14 4552303 Qxc6 dxc6 Rc8
>00:00:18.8 -2,36 15 8193181 Qxc6 dxc6 Kg8 Bd5+ Kh7 Be6 Bc7 Ra7 Bb6 Rf7 Ra8 c7
>Ra1+ Ke2 Ra2+ Kf3 Bxc7 Rxc7
>00:02:37.0 -2,20 15 68200230 Qe3 Qd7+ Kg8 Qxd6 Rf8 Qe6+ Kh7 Qe7 Rg8 Bf3 Qxc3 Qe6
>Qxb4 d6 Qb1+ Kf2 Qc2+ Be2
>00:04:10.3 -1,30 16 110849691 Qe3
>00:04:20.5 -1,30 17 115758982 Qe3
>00:11:53.6 -1,44 17 336003972 Qe3 Qxd6 Re8 h4 Re7 Bf3 Qc1+ Kf2 Qd2+ Kg3 Qe1+ Kg4
>e4 Qc5 exf3 Re6 Rxe6 fxe6+
>00:18:17.0 -1,48 18 525918038 Qe3 Qxd6 Re8 h4 h5 Bf3 Qc1+ Kf2 Qd2+ Be2 Qf4+ Kg1
>Qe3+ Kh2 Qf4+ Kh3 Qxf5+ Kh2
>00:29:14.0 -1,48 19 860756587 Qe3 Qxd6 Re8 h4 h5 Bf3 Qc1+ Kf2 Qd2+ Be2 Qf4+ Kg1
>Qe3+ Kh2 Qf4+ Kh3 Qxf5+ Kh2
>01:07:39.2 -1,54 20 2044093935 Qe3 Qxd6 Re8 h4 h5 Bf3 Qc1+ Kf2 Qd2+ Be2 Qf4+ Kg1
>Qe3+ Kh2 Qf4+ Kh3 Qxf5+ Kh2
>02:32:08.3 -1,54 21 432997679 Qe3 Qxd6 Re8 h4 h5
>
>(nothing more for the next 10 hours or so)
>
>--
>GCP
The problem with the node counter is related to the GUI. The engine provides
both a 32 bits and a 64 bits version of the counter since version 14, but I
think the GUI still uses the 64 bits counter.
Short PVs happen all the time with the Tiger engine (and many other engines as
well) due to specific search algorithms and/or hash table hits. I don't see a
problem here.
If the program takes time to show the next iteration, it might just be because
it has just seen something important or unexpected. After all, that's exactly
the reason why you give it a lot of time to think, isn't it?
Chess Tiger has been tested on overnight analysis, and as far as I know works
fine.
Christophe
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.