Code: Select all
move divide(26)
----------------------------
12-16: 111362678435231752
11-16: 112590257768420515
11-15: 101352649993886926
10-15: 100552646996749293
10-14: 86312674861234785
9-14: 103811787278058952
9-13: 136189763354484914
----------------------------
perft(26) 752172458688067137
Hi AartAartBik wrote:I made some improvements to my distributed implementation for computing perft and divide numbers for 8x8 checkers that enabled me to go even deeper, and now computed depth 26 from the initial position, shown below. As reported earlier, the numbers were computed on a cluster of machines, further optimized with a "hard collision"-free transposition table as well as bulk counting. The move generator does not eliminate duplicate captures (viz. the situation where a king can capture the same pieces in different directions; a situation that starts to occur at depth 12 and up).
Code: Select all
move divide(26) ---------------------------- 12-16: 111362678435231752 11-16: 112590257768420515 11-15: 101352649993886926 10-15: 100552646996749293 10-14: 86312674861234785 9-14: 103811787278058952 9-13: 136189763354484914 ---------------------------- perft(26) 752172458688067137
But congratulations are more in order here! Great stuff. If I ever try my hand at perft() again, I would probably have to do a generalized divide(depth, 5) (i.e. divide the computation into subtrees 5 plies from the root) in order to be able to save intermediate results and guard against power failures. [Note: the familiar perft(depth) and divide(depth) are divide(depth, 0) and divide(depth, 1) in this notation].What else did you expect after your previous "smallest possible increments" comments? But seriously, I am having fun with these computations, and hope others still enjoy it too. Yesterday I also updated the A133046 entry with proper credit to you and Murray. In any case, I am hoping for a confirmation of the higher depths, or perhaps even a challenge for perft(27)?Rein Halbersma wrote:My initial thought was: "arggggggggggggggh".
Ha! Good to see that your inner sportsman was triggered by my needlingAartBik wrote:What else did you expect after your previous "smallest possible increments" comments? But seriously, I am having fun with these computations, and hope others still enjoy it too. Yesterday I also updated the A133046 entry with proper credit to you and Murray. In any case, I am hoping for a confirmation of the higher depths, or perhaps even a challenge for perft(27)?Rein Halbersma wrote:My initial thought was: "arggggggggggggggh".
I'm not embarking on any kind of perft compuation with my current P4, but I'm in the process of ordering a Xeon E5-1650 (6-core) that should be able to at least break my personal best of perft(22)... And thanks for crediting me at OEIS.Looking forward to that!Rein Halbersma wrote:I'm not embarking on any kind of perft compuation with my current P4, but I'm in the process of ordering a Xeon E5-1650 (6-core) that should be able to at least break my personal best of perft(22).
