Author: Robert Hyatt
Date: 05:36:00 11/22/05
Go up one level in this thread
On November 21, 2005 at 09:04:22, Andreas Guettinger wrote: >On November 21, 2005 at 04:36:19, enrico carrisco wrote: > >>On November 21, 2005 at 04:34:26, Andreas Guettinger wrote: >> >>>On November 20, 2005 at 22:42:02, enrico carrisco wrote: >>> >>>>All find mate -- some do it instantly, but none take over 20 seconds... >>>> >>>>Now let's use Fruit 2.2.1 in the same conditions (no egtb): >>>> >>>>8/2k5/8/3R4/8/8/5K2/8 w - - 0 1 >>>> >>>>Analysis by Fruit 2.2.1: >>>> >>>>1.Ke3 >>>> +- (5.45) Depth: 1/1 00:00:00 >>>>1.Ke3 Kc6 >>>> +- (5.57) Depth: 2/2 00:00:00 >>>>1.Ke3 Kc6 2.Ke4 >>>> +- (5.45) Depth: 3/4 00:00:00 >>>>1.Kg3 Kb7 2.Kf4 >>>> +- (5.55) Depth: 3/6 00:00:00 >>>>1.Kg3 Kb7 2.Kf4 Kc6 >>>> +- (5.58) Depth: 4/6 00:00:00 >>>>1.Ke3 Kc6 2.Ke4 Kc7 >>>> +- (5.76) Depth: 4/6 00:00:00 >>>>1.Ke3 Kc6 2.Ke4 Kc7 3.Rc5+ Kd6 >>>> +- (5.57) Depth: 5/8 00:00:00 >>>>1.Ke3 Kc6 2.Kd4 Kb6 3.Rc5 Kb7 >>>> +- (5.85) Depth: 6/10 00:00:00 13kN >>>>1.Ke3 Kc6 2.Kd4 Kb6 3.Rd6+ Kb5 4.Rg6 Kb4 5.Rb6+ Ka4 >>>> +- (5.89) Depth: 7/11 00:00:00 58kN >>>>1.Ke3 Kc6 2.Kd4 Kb6 3.Rd6+ Kb5 4.Rg6 Kb4 5.Rb6+ Ka5 >>>> +- (5.89) Depth: 8/13 00:00:01 238kN >>>>1.Ke3 Kc6 2.Kd4 Kb6 3.Rd7 Kb5 4.Rd6 Kb4 5.Rb6+ Ka5 >>>> +- (5.89) Depth: 9/14 00:00:01 972kN >>>>1.Ke3 Kc6 2.Kd4 Kb6 3.Ke5 Kc7 4.Rc5+ Kd7 5.Rc1 Ke7 6.Rc7+ Kd8 >>>> +- (6.25) Depth: 10/18 00:00:01 3671kN >>>>1.Ke3 Kc6 2.Kd4 Kb6 3.Ke5 Kc7 4.Rb5 Kd7 5.Rc5 Ke7 6.Rc7+ Kd8 >>>> +- (6.25) Depth: 11/19 00:00:05 13571kN >>>>1.Rh5 Kd6 2.Ke3 Ke6 3.Kd4 Kd6 4.Rh6+ Ke7 5.Rc6 Ke8 6.Rc7 Kd8 >>>> +- (6.26) Depth: 11/20 00:00:18 48869kN >>>>1.Rh5 Kd6 2.Ke2 Kd7 3.Kd3 Kc6 4.Rh6+ Kc7 5.Rg6 Kb8 6.Rg7 Kc8 7.Kd4 >>>> +- (6.14) Depth: 12/20 00:00:28 77398kN >>>>1.Ke3 Kc6 2.Ke4 Kc7 3.Ke5 Kc6 4.Ke6 Kc7 5.Rd6 Kc8 6.Rd7 Kb8 >>>> +- (6.33) Depth: 12/20 00:00:30 83851kN >>>>1.Ke3 Kc6 2.Kd4 Kb6 3.Rc5 Kb7 4.Ke5 Ka7 5.Rc6 Kb8 6.Rc4 Ka8 7.Rc7 >>>> +- (6.45) Depth: 13/22 00:01:43 279150kN >>>> >>>> >>>>Analysis stopped after 10 minutes -- all analysis on an A64 @ 2.7GHz with 512MB >>>>hash (except CT 14 due to hash size control limitations.) >>>> >>>>Please -- will the die-hard "EGTB files are a waste" believers please stand-up. >>>> >>>>Regards, >>>> >>>>Enrico. >>> >>>Deph 13 in 1:43 min in an endgame is indeed a bit hard to believe. >>>Smells like bug. >>> >>>A similar behaviour has been reported previously. I think it seems to occur >>>whether EGTB are on or off, if they are on the EGTB are suddenly not accessed >>>anymore in a subsequent analysis with very slow search. >>> >>> >>>regards >>>Andy >> >> >>Greetings. >> >>Yes, I missed that earlier thread (perhaps weeks ago?) and we discovered that >>the hash wasn't clearing further below (at least that is the best guess to >>explain it.) >> >>Regards, >> >>-elc. > >It was maybe 3-4 weeks ago. I wonder if somebody actually filed the bug to the >Fruit team. >There has too be more than the hash clearing, or how else can the 13 ply depth >and the very slow search be explained? > >- Andy I don't see why the hash would need to be cleared, so long as "adjusted" mate scores are stored. I do this to _avoid_ EGTB probes when possible, as the hash table is far faster than searching the EGTB cache for a hit.
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.