Author: Stuart Cracraft
Date: 13:24:41 09/17/04
Go up one level in this thread
Bitboards and parallel search seem like apples and oranges. I don't see why bitboards would affect whether or not a parallel search is successful or not. Depends on your splitting algorithm, communications overhead, and a lot more. Of more interest is what is your design for those and do you scale favorably above just a couple processors? Is your algorithm any relation to DTS or how Bob handles it? I recall papers he wrote on this (I think) years ago with very high % success rate on multi-processor utilization but think it fell off at 4 or 8 processors. I forget the reason why. What was the success rate of Deep Thought on its larger parallelization? How about Star Socrates or others? With the dual-core cpu's imminent (we had a big presentation by IBM at my company touting this stuff), we can all expect a nice big jump in depth and NPS soon for some extra programming. But the question is, what's the best algorithm to use and how does it scale? Stuart On September 17, 2004 at 14:33:41, Dann Corbit wrote: >http://www.imm.dtu.dk/pubdb/views/publication_details.php?id=3267 > >I got ~96% efficiency, going from 1 thread to 2 threads on a dual CPU machine. >1172442 / 599317 = 1.9563 > > >WindowsNT Mgibian 0 5 x86 >Reading 'WAC230.EPD' >Thread 0 created (main thread) >Threads started: 1 >Setting up FEN: 2b5/1r6/2kBp1p1/p2pP1P1/2pP4/1pP3K1/1R3P2/8 b - - bm Rb4; id >"WAC.230"; >time nodes depth score pv >00:00:00.00 1 01|01|01 +1.095 Kd7 >00:00:00.00 16 --------------------------------------------------- > -- >00:00:00.00 104 02|02|03 +0.595 Kd7 Kg2 >00:00:00.00 149 --------------------------------------------------- > ++ >00:00:00.00 293 03|03|03 +1.395 Kd7 Kg2 Ke8 >00:00:00.00 323 --------------------------------------------------- > -- >00:00:00.00 579 04|04|05 +0.895 Kd7 Kg2 Ke8 Kg1 >00:00:00.00 745 --------------------------------------------------- >00:00:00.00 1060 05|05|06 +0.955 Kd7 Kg2 Ke8 Kg1 Bd7 >00:00:00.00 1573 --------------------------------------------------- >00:00:00.00 2913 06|06|07 +0.940 Kd7 Kg2 Ke8 Kg1 Bd7 f4 >00:00:00.00 3350 --------------------------------------------------- >00:00:00.01 4919 07|07|08 +0.950 Kd7 Kg2 Ke8 Kg1 Bd7 f4 Bc6 > ++ >00:00:00.03 19587 07|07|09 +1.380 Rf7 Kg2 Kb7 Rd2 Rf5 Kg1 Rxg5 >00:00:00.03 20127 --------------------------------------------------- >00:00:00.07 46353 08|09|11 +1.090 Rf7 Kg2 Kd7 Kg1 Rf5 f4 Rxf4 Rd2 >00:00:00.07 49469 --------------------------------------------------- >00:00:00.20 122131 09|12|13 +0.965 Rf7 Rb1 Kd7 Ra1 b2 Rb1 Ke8 Rxb2 Bd7 >00:00:00.28 161342 --------------------------------------------------- >00:00:00.57 340212 10|13|16 +0.950 Rf7 Rb1 Bb7 Ra1 Kd7 Rxa5 Ke8 f4 b2 Rb5 >00:00:00.84 512496 10|13|16 +1.060 Kd7 f4 Ke8 Kh2 Rh7 Kg1 Rh3 Bc7 Rg3 Kh1 >Rxc3 Bxa5 >00:00:00.84 514536 --------------------------------------------------- >00:00:00.95 589447 11|14|16 +0.955 Kd7 Kg2 Ke8 Rb1 Bd7 Kg1 Bc6 f4 Rf7 Rf1 >Rd7 >00:00:01.54 922874 11|14|17 +0.975 Rf7 Rb1 Rf5 Be7 Kd7 Bf6 Ke8 Ra1 Bd7 >Rxa5 Bc6 >00:00:01.62 970894 --------------------------------------------------- > > time used: 00:00:01.62 > nps: 599317 nodes: 970894 Q-nodes: 95579 (9%) > move order: 89% > hash probes: 338964 hash hits: 199491 (58%) > hash usage: 20% > >My move: Rf7 > > > >WindowsNT Mgibian 0 5 x86 >Reading 'WAC230.EPD' >Thread 0 created (main thread) >Creating worker thread 1 >Threads started: 2 >Setting up FEN: 2b5/1r6/2kBp1p1/p2pP1P1/2pP4/1pP3K1/1R3P2/8 b - - bm Rb4; id >"WAC.230"; >time nodes depth score pv >00:00:00.00 1 01|01|01 +1.095 Kd7 >00:00:00.00 16 --------------------------------------------------- > -- >00:00:00.00 104 02|02|03 +0.595 Kd7 Kg2 >00:00:00.00 149 --------------------------------------------------- > ++ >00:00:00.00 293 03|03|03 +1.395 Kd7 Kg2 Ke8 >00:00:00.00 323 --------------------------------------------------- > -- >00:00:00.00 580 04|04|05 +0.895 Kd7 Kg2 Ke8 Kg1 >00:00:00.00 746 --------------------------------------------------- >00:00:00.00 1063 05|05|06 +0.955 Kd7 Kg2 Ke8 Kg1 Bd7 >00:00:00.00 1585 --------------------------------------------------- >00:00:00.00 2948 06|06|07 +0.940 Kd7 Kg2 Ke8 Kg1 Bd7 f4 >00:00:00.00 3385 --------------------------------------------------- >00:00:00.00 5013 07|07|08 +0.950 Kd7 Kg2 Ke8 Kg1 Bd7 f4 Bc6 > ++ >00:00:00.06 22019 07|07|09 +1.380 Rf7 Kg2 Kb7 Rd2 Rf5 Kg1 Rxg5 >00:00:00.06 22573 --------------------------------------------------- >00:00:00.08 50490 08|09|11 +1.090 Rf7 Kg2 Kd7 Kg1 Rf5 f4 Rxf4 Rd2 >00:00:00.09 53702 --------------------------------------------------- >00:00:00.17 147570 09|12|13 +0.965 Rf7 Rb1 Kd7 Ra1 b2 Rb1 Ke8 Rxb2 Bd7 >00:00:00.20 190707 --------------------------------------------------- >00:00:00.36 374858 10|13|16 +0.950 Rf7 Rb1 Bb7 Ra1 Kd7 Rxa5 Ke8 >00:00:00.52 575781 10|13|16 +1.060 Kd7 f4 Ke8 Kh2 Rh7 Kg1 Rh3 Bc7 Rg3 Kh1 >Rxc3 Bxa5 >00:00:00.52 577991 --------------------------------------------------- >00:00:00.63 711651 11|15|16 +0.960 Kd7 Kg2 Ke8 Rb1 Bd7 Kg1 Bc6 f4 b2 >00:00:01.03 1202303 11|15|17 +0.975 Rf7 Rb1 Rf5 Be7 Kd7 Bf6 Ke8 Ra1 Bd7 >Rxa5 Bc6 >00:00:01.06 1242789 --------------------------------------------------- > > time used: 00:00:01.06 > nps: 1172442 nodes: 1242789 Q-nodes: 122113 (9%) > move order: 90% > hash probes: 457029 hash hits: 269083 (58%) > hash usage: 25% > >My move: Rf7
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.