Author: Tom Kerrigan
Date: 01:20:22 06/27/03
Go up one level in this thread
On June 26, 2003 at 19:11:13, Steven J. Brann wrote: >On June 26, 2003 at 13:32:36, John Merlino wrote: > >>On June 26, 2003 at 11:18:26, Igor Gorelikov wrote: >> >>>One Test position >>>================== >>> >>>There is a section of everyday's puzzles on my site: >>>http://www.digichess.gr/infiniteloop/home.php >>> >>>All puzzles are taken from engine to engine games of Infinite Loop. >>> >>>Prior to publish a puzzle on my site I always offer Crafty 19.3 to solve it. >>> >>>I was amazed that today's puzzle was hard to solve even for Crafty so it >>>can be a good test position for any chess engine. >>> >>>What Phalanx XXII (engine number 10 among free WB engines) plays here as >>>white (and find it at once)? >>> >>>[D] r1bq3r/ppp3pp/3k4/3p4/3nP1Q1/8/PP4PP/RN3RK1 w - - 0 17 >>> >>>PS: Yace can find it after some time too. >>> >>>Best regards, >>>Igor Gorelikov >> >>On a moderately busy (since I'm at work) and humble P3-750, it takes Chessmaster >>9000 (default personality) almost six minutes to produce a score > 1.0, even >>though The King chooses e5+ instantly. So, on today's top hardware, it will >>definitely produce a +1 score in under two minutes: >> >>Time Depth Score Positions Moves >>0:00 1/3 0.64 3599 17.e5+ Kxe5 18.Qxg7+ Kd6 19.Qxd4 >> Rf8 >>0:00 1/4 1.21 10186 17.e5+ Kxe5 18.Qxg7+ Ke6 19.Qxd4 >> Rf8 20.Nc3 Rxf1+ 21.Rxf1 >>0:00 1/5 0.89 37549 17.e5+ Kxe5 18.Qxg7+ Kd6 19.Qxd4 >> c5 20.Qf4+ Kc6 21.Nc3 >>0:01 1/6 0.83 117405 17.e5+ Kxe5 18.Qxg7+ Kd6 19.Qxd4 >> Rf8 20.Nc3 Rxf1+ 21.Rxf1 Be6 >>0:03 1/7 0.77 381074 17.e5+ Kxe5 18.Qf4+ Ke6 19.Qxd4 >> Rf8 20.Nc3 Rf6 21.Rae1+ Kf7 22.Nxd5 >> Rxf1+ 23.Kxf1 >>0:13 1/8 0.92 1320923 17.e5+ Kxe5 18.Qxg7+ Kd6 19.Qxd4 >> c5 20.Qf4+ Kc6 21.Nc3 b5 22.Qf7 Qd6 >>0:34 1/9 0.84 3591060 17.e5+ Kxe5 18.Qxg7+ Kd6 19.Qxd4 >> Be6 20.Nc3 Kd7 21.Rae1 c6 22.Qg7+ >> Qe7 >>1:41 1/10 0.94 10629633 17.e5+ Kxe5 18.Qxg7+ Kd6 19.Qxd4 >> Be6 20.Nc3 Kd7 21.Rae1 Rf8 22.Rxf8 >> Qxf8 23.Nxd5 Bxd5 24.Qxd5+ Kc8 >>5:49 2/11 1.15 34058081 17.e5+ Kxe5 18.Qxg7+ Kd6 19.Qxd4 >> Be6 20.Nc3 Kd7 21.Rae1 Qg8 22.Ne4 >> Kc8 23.Nd6+ cxd6 24.Rxe6 >> >>jm > >John, You're right... under two minutes <S>. I have a new Win XP 3.0G P4 with >hyperthreading which, for Chessmaster is the best thing since sliced bread. Not >only is it fast but In the task manager shows only 50% of the CPU utilized. I >can run other applications concurrently with CM while mentor lines are >calculating and the other applications don't drag a bit. Task Manager isn't being completely honest in this situation... you have two logical processors and only one thread is running, so TM sees that one of the logical processors (50%) is free... in fact, that one thread is taking up ~100% of your CPU... if you run another CPU-intensive program while running ChessMaster, both will go much slower than if you ran them separately. >Chessmaster when "thinking" brought my 1.9G P4 to its knees. If the screen >saver happened to kick in it would take over 2 minutes after wiggling the mouse >to get the screen back. Right now while mentor (CM default with SS=12) >evaluates this position, I'm writing this, polling for email, have two instant >messengers going and am checking spelling once in awhile using Word. Granted, >the other applications don't take much CPU, but they're certainly not dragged >down while CM is having a deep think. Incredible and a pleasant surprise. That's because of a design flaw with the way Windows schedules processes. UI intensive programs like Word and Outlook yield control of the CPU frequently while in the middle of doing stuff. Let's say you're doing something in Word that takes it 20ms. It computes for 1ms, then yields control back to the OS. Then it computes for another ms, and yields control again. If you're not doing anything CPU-intensive, this is fine, because when it yields control, it gets control back in a tiny fraction of a ms. If you do this while running ChessMaster in the background, Word will compute for 1ms, yield control, and then ChessMaster will compute for 20ms. Word will compute for another ms, then CM will compute for 20ms. So your operation in Word that usually takes 20ms all of a sudden takes 420ms--half a second. 20 times slower. When you have two processors (or logical processors), ChessMaster gets its own processor to chug away on and then when Word yields control on the other processor, nothing else is running on that processor to suck up time, so Word gets control back immediately and is consequently fast again. -Tom
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.