After some more optimizations my Java perft now runs much faster than before, it is now nearly half as fast as the C code. I would have expected much less from Java and am pleasantly surprised. I'm quite happy already with the Java engine, so now for the hard part - creating an interface....
perft 1 time 1 positions 7 kN/s 7
perft 2 time 1 positions 49 kN/s 49
perft 3 time 1 positions 302 kN/s 302
perft 4 time 1 positions 1469 kN/s 1469
perft 5 time 1 positions 7361 kN/s 7361
perft 6 time 15 positions 36768 kN/s 2451
perft 7 time 1 positions 179740 kN/s 179740
perft 8 time 63 positions 845931 kN/s 13427
perft 9 time 250 positions 3963680 kN/s 15854
perft 10 time 1156 positions 18391564 kN/s 15909
perft 11 time 5437 positions 85242128 kN/s 15678
perft 12 time 24954 positions 388623673 kN/s 15573
Subscribe to:
Post Comments (Atom)
2 comments:
I've let my own primitive checkers program (named "milhouse") lie dormant for over a year, but your awakening has reminded me to dust it off again. Your perft numbers are apparently quite helpful, and have told me that I very likely have a bug, since these are the numbers I get:
depth 1 positions 7
depth 2 positions 56
depth 3 positions 358
depth 4 positions 1827
depth 5 positions 9188
depth 6 positions 45956
depth 7 positions 225696
depth 8 positions 1071627
depth 9 positions 5035307
depth 10 positions 23426871
depth 11 positions 108668999
depth 12 positions 497292672
Not sure what's going on here, but it should be obvious when I print out all the level two positions...
Doh. Stupid mistake. I counted all nodes, not just leaf nodes. I instrumented my code (C) and ran it, yielding:
running perft performance/accuracy test.
depth 1 positions 7 (0.000 seconds, 699 KN/s)
depth 2 positions 49 (0.000 seconds, 3805 KN/s)
depth 3 positions 302 (0.000 seconds, 7771 KN/s)
depth 4 positions 1469 (0.000 seconds, 10568 KN/s)
depth 5 positions 7361 (0.001 seconds, 12050 KN/s)
depth 6 positions 36768 (0.003 seconds, 11566 KN/s)
depth 7 positions 179740 (0.015 seconds, 11647 KN/s)
depth 8 positions 845931 (0.077 seconds, 11045 KN/s)
depth 9 positions 3963680 (0.354 seconds, 11204 KN/s)
depth 10 positions 18391564 (1.375 seconds, 13373 KN/s)
depth 11 positions 85242128 (3.601 seconds, 23672 KN/s)
depth 12 positions 388623673 (16.578 seconds, 23441 KN/s)
Post a Comment