Author: Robert Hyatt
Date: 13:46:24 06/29/99
Go up one level in this thread
On June 29, 1999 at 14:04:34, Vincent Diepeveen wrote: >On June 29, 1999 at 13:57:06, Robert Hyatt wrote: > >>On June 29, 1999 at 11:09:41, Vincent Diepeveen wrote: >> >>>Hello, >>> >>>Hereby the parallel testset as used by cray blitz. >>>It's move 9 from black to move till move 32 from black to move >>> >>>Average speedup of cray blitz: 3.7 >> >> >>beware. This is _not_ a test set. It was a complete game fragment, that I >>contrived CB to follow. IE I played this real game at an ACM event vs >>MchessPro, using 16 cpus. I then replayed the game with 8 processors, then >>with 4, then with 2, and finally with 1, but doing the normal pondering stuff >>and everything. IE I pretended to play the game with 1,2,3,8 and 16 processors, >>but forced the search to reach the same depth no matter what, which took forever >>for 1 cpu. I saved the hash table between moves, passed the PV along, and so >>forth. > >Hmm. Using a testset i don't save hashtable. That indeed makes up for >a big extra speedup. But no problems here. I'll simply let it analyze a little >longer then. It's just the compare between some programs. > >I'm missing other volunteers doing this by the way > >>Using this as a 'test suite' isn't a good comparison, because my test approach >>was more favorable to CB than just stuffing in N positions and searching each >>from scratch (say using "test command" in crafty)... > >Does this mean that your official publication in ICCA wasn't >as officially correct as it should be? No. The ICCA publication explains _precisely_ how I ran the test, which is just as I explained it above. I even explained _why_ I ran a "game" rather than a set of "positions"... > > >>> >>>Number of nodes visited to come to this 3.7 speedup: >>>about 100 million a position, which at the cray took at >>>16 processors between 84 and 814 seconds. >>> >>>Report to be continued soon! >>> >>>r2qkbnr/ppp2p1p/2n5/3P4/2BP1pb1/2N2p2/PPPQ2PP/R1B2RK1 b Qkq - ; move 9 black >>>a1a1 >>>r2qkbnr/ppp2p1p/8/nB1P4/3P1pb1/2N2p2/PPPQ2PP/R1B2RK1 b Qkq - ; move 10 of black >>>a1a1 >>>r2qkbnr/pp3p1p/2p5/nB1P4/3P1Qb1/2N2p2/PPP3PP/R1B2RK1 b Qkq - ; move 11 of black >>>a1a1 >>>r2qkb1r/pp3p1p/2p2n2/nB1P4/3P1Qb1/2N2p2/PPP3PP/R1B1R1K1 b Qkq - ; move 12 of >>>black >>a1a1 >>>r2q1b1r/pp1k1p1p/2P2n2/nB6/3P1Qb1/2N2p2/PPP3PP/R1B1R1K1 b - - ; move 13 of black >>>a1a1 >>>r2q1b1r/p2k1p1p/2p2n2/nB6/3PNQb1/5p2/PPP3PP/R1B1R1K1 b - - ; move 14 of black >>>a1a1 >>>r2q1b1r/p2k1p1p/2p5/nB6/3Pn1Q1/5p2/PPP3PP/R1B1R1K1 b - - ; move 15 of black >>>a1a1 >>>r2q1b1r/p1k2p1p/2p5/nB6/3PR1Q1/5p2/PPP3PP/R1B3K1 b - - ; move 16 of black >>>a1a1 >>>r2q1b1r/p1k2p1p/8/np6/3PR3/5Q2/PPP3PP/R1B3K1 b - - ; move 17 of black >>>a1a1 >>>r4b1r/p1kq1p1p/8/np6/3P1R2/5Q2/PPP3PP/R1B3K1 b - - ; move 18 of black >>>a1a1 >>>r6r/p1kqbR1p/8/np6/3P4/5Q2/PPP3PP/R1B3K1 b - - ; move 19 of black >>>a1a1 >>>5r1r/p1kqbR1p/8/np6/3P1B2/5Q2/PPP3PP/R5K1 b - - ; move 20 of black >>>a1a1 >>>5r1r/p2qbR1p/1k6/np2B3/3P4/5Q2/PPP3PP/R5K1 b - - ; move 21 of black >>>a1a1 >>>5rr1/p2qbR1p/1k6/np2B3/3P4/2P2Q2/PP4PP/R5K1 b - - ; move 22 of black >>>a1a1 >>>5rr1/p2qbR1p/1kn5/1p2B3/3P4/2P2Q2/PP4PP/4R1K1 b - - ; move 23 of black >>>a1a1 >>>4qRr1/p3b2p/1kn5/1p2B3/3P4/2P2Q2/PP4PP/4R1K1 b - - ; move 24 of black >>>a1a1 >>>5qr1/p3b2p/1kn5/1p1QB3/3P4/2P5/PP4PP/4R1K1 b - - ; move 25 of black >>>a1a1 >>>5q2/p3b2p/1kn5/1p1QB1r1/P2P4/2P5/1P4PP/4R1K1 b - - ; move 26 of black >>>a1a1 >>>5q2/p3b2p/1kn5/3QB1r1/p1PP4/8/1P4PP/4R1K1 b - - ; move 27 of black >>>a1a1 >>>5q2/p3b2p/1k6/3QR1r1/p1PP4/8/1P4PP/6K1 b - - ; move 28 of black >>>a1a1 >>>5q2/p3b2p/1k6/4Q3/p1PP4/8/1P4PP/6K1 b - - ; move 29 of black >>>a1a1 >>>3q4/p3b2p/1k6/2P1Q3/p2P4/8/1P4PP/6K1 b - - ; move 30 of black >>>a1a1 >>>3q4/p3b2p/8/1kP5/p2P4/8/1P2Q1PP/6K1 b - - ; move 31 of black >>>a1a1 >>>3q4/p3b2p/8/2P5/pk1P4/3Q4/1P4PP/6K1 b - - ; move 32 of black >>>a1a1
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.