Author: Tim Foden
Date: 23:47:18 08/10/00
Go up one level in this thread
On August 10, 2000 at 17:42:56, José de Jesús García Ruvalcaba wrote: >On August 10, 2000 at 17:23:32, Tim Foden wrote: > >>On August 10, 2000 at 16:17:47, Christoph Fieberg wrote: >> >>>Unbelievable, but true: >>>I composed a position which is mate in 1 and can easily found, but Fritz6 on >>>Pentium III, 500 Mhz and 32 MB Hash needs more than 1 hour to show it (excatly >>>1:03:17)!!! >>>How do other computers react? >>> >>>The position is >>>FEN: [D]8/8/pppppppK/NBBR1NRp/nbbrqnrP/PPPPPPPk/8/Q7 w - - 0 1 >> >>Green Light takes just over 29 seconds on my K6/2 333Mhz: >> >>------------------ >> fen 8/8/pppppppK/NBBR1NRp/nbbrqnrP/PPPPPPPk/8/Q7 w - - 0 1 >> Game finished test: 0 >> time limit 178.00 >> Game stage: Opening >> Current eval: -0.21 >> Ply CE Time Nodes Principal variation >> 1 ++ 23 18913 Nxc6 (a=-0.71 b=0.29 e=0.29) >> 1 35 107 94418 Nxc6 >> 1 245 221 194150 Nxc4 >> 1 431 249 225981 Nxd4 >> 1 488 250 227290 Rxd4 >> 1 521 1771 1641120 fxe4 >> 1 719 2891 2779247 dxe4 >> 1 32099 2907 2795807 Qh1# >> 1 32099 2908 2795811 Qh1# >> Max eval diff 1 2.39, 2 0.98 >> local: t=29.08 nps=96135.4 n=2795811 (f=3 q=2795808) >> total: t=7:30 nps=110487.0 n=49749871 >> extensions: check=2 recapture=0 pawn push=0 1-reply=0 >> q-moves: generated=9774356 considered=5271969 made/unmade=2795759 >> !Qa1-h1 32099 >>move Qa1-h1 >>-------------------- >> >>The major reason for the time usage is the quiescence search, which it the bit >>that searches until all potential capture gains have been found. >> >>If you look carfully... GLC has only searched full alpha-beta 3 nodes!! >>But it searched 2,795,808 quiecence nodes. >> >>Cheers, Tim. > >Is Green Light your program? Yes. >It shows six different PV's I wonder how it >considered only three nodes in full. >José. As the program only searched 1 ply it was into quiescence search under the root node. At the root of the search we make all moves, but then the program only calls the quiescence search to make its descisions. The 2 other nodes searched we due to check extensions at the root level (for Qf1+ and Qh1#). The PV's shown are as the search finds better moves at the root. I should really add some initial root move ordering, so that, say, checks are done first and captures are ordered to capture the highest value piece first. This would have cut out most of the other PV's in this example. Cheers, Tim.
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.