Author: Uri Blass
Date: 06:48:01 01/08/03
Go up one level in this thread
On January 08, 2003 at 09:13:38, Peter Fendrich wrote: >On January 08, 2003 at 08:58:08, Uri Blass wrote: > >>On January 08, 2003 at 08:42:20, Peter Fendrich wrote: >> >>>On January 08, 2003 at 04:51:15, Russell Reagan wrote: >>> >>>>On January 08, 2003 at 04:37:09, Peter Fendrich wrote: >>>> >>>>>You should know by now...-:) >>>>>Did you find out what your perft problem was? >>>> >>>>No, but it's fixed :) I re-worked the way I handled castling rights and made >>>>sure I was updating them correctly in special cases (like when a rook is >>>>captured, remove its castling rights), and it worked, so it must have been >>>>something with the castling rights, but I never nailed down exactly what it was. >>>>I changed several things (mostly those special cases) and it worked, so it must >>>>have been one of those cases (or several) that got fixed. Thanks for your help >>>>BTW ;-) >>> >>>In that case, you're ready for the "final" test! >>>[D]r3k2r/p1ppqpb1/bn2pnp1/3PN3/1p2P3/2N2Q1p/PPPBBPPP/R3K2R w KQkq - >>>1: 48 >>>2: 2039 >>>3: 97862 >>>4: 4085603 >>>5: 193690690 >>>It tests all the rules in chess except for a few situations, like this one: >>>[D]k7/8/8/2r1Pp1K/8/8/8/8 w - f6 >>>exf6 e.p. is invalid >>>1: 5 >>>2: 75 >>>3: 471 >>>4: 7883 >>>5: 50052 >>> >>>Regards, >>>Peter >> >>Note that the only way to use movei for that test in this case is the follwoing >>way. >> >>1)Give it >> >>k7/5p2/8/2r1P2K/8/8/8/8 b - - 0 1 >> >>2)play f7f5 in edit mode >>3)ask it for perft 5. >> >>perft 6=860303 >>perft 7=5650590 >>perft 8=98654119 >>perft 9=681267616 >> >>If I have no bugs >> >>Other people may calculate perft 10 and perft 11 to check if movei is right. >> >>Uri > >I'm sitting on a 300 MHz computer right now and even perft 9 would take forever >but perft 7 gives the same result between Terra and Movei. I can't imagine that >anything new will appear in this particular position after perft 9 except >possibly overflows... > >Peter perft 9 took me few minutes on 1000 Mhz. Perft 10 may take more time but it is doable and movei knows about numbers that are bigger than 2^32 for the perft function so overflow is not going to happen. Uri
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.