Page 1 of 1
Looking for Nalimov 5-men tbgen source
Posted:
05 May 2007, 23:02
by Sven Schüle
Hi all,
can anyone post a really working link to the sources (TBGEN.ZIP) of the old 5-men Nalimov tablebase generator (provided Eugene Nalimov still allows reading them)?
I mean the generator itself, not only the probing code (which is indirectly available through Crafty and others).
IIRC they had been online a while ago at least at Robert Hyatt's page but now I don't find them anymore. Tried also many Google links, everything seems dead.
Please, no "Why do you need the generator, you can download all (3-4-5-)6-men TBs from XY" answers, because my main intention is not having the TBs but looking at the generator source itself. I already had written an own 4-men generator but it was very inefficient, and last year I lost the sources unfortunately. Now I want to restart somehow (not by "cloning", of course), just for my own fun at the moment.
Any help would be appreciated. You may also send me an email (see bottom of the "Surprise" homepage).
Sven
Re: Looking for Nalimov 5-men tbgen source
Posted:
06 May 2007, 01:02
by Tony Mokonen
The Yahoo! chess-engines group has the tbgen source, in the Files section. You have to sign up with Yahoo! first and join the chess-engines group before you can access the file area.
Tony
Re: Looking for Nalimov 5-men tbgen source
Posted:
06 May 2007, 08:45
by Jim Ablett
Re: Looking for Nalimov 5-men tbgen source
Posted:
06 May 2007, 20:18
by Sven Schüle
Thanks Tony and Jim, that's what I looked for!
Sven
Re: Looking for Nalimov 5-men tbgen source
Posted:
07 May 2007, 14:33
by Sven Schüle
Hmmmm ... @Jim: this Tbgen.zip seems to be incomplete, at least tbdecode.c is missing (it is included from tbindex.cpp which in turn is included from Tbgen.cpp).
@Tony: I can't find a Yahoo! group named "chess-engines". I logged in with my Yahoo ID and searched for groups with "chess-engines", "chess engines", or just "chess", but I only found "chessdatabase", "Chess_Engines_Testing", "Chess_Library", and "Chess_Improvement". Did you mean one of these? Or what am I doing wrong?
Any ideas?
Sven
Re: Looking for Nalimov 5-men tbgen source
Posted:
07 May 2007, 15:43
by Volker Pittlik
Sven Schüle wrote:...I can't find a Yahoo! group named "chess-engines". ...
Any ideas?
http://groups.yahoo.com/group/chess-engines/
Volker
Re: Looking for Nalimov 5-men tbgen source
Posted:
07 May 2007, 15:59
by Sven Schüle
Now I got a compiling version, thanks Volker!
The "chess-engines group" version seems to be the very old one from 1998 but I think it's sufficient for now to try to understand the basics.
Nevertheless I would like to know whether someone has a newer version which is complete (5- or 6-men, does not matter).
Regards,
Sven
Re: Looking for Nalimov 5-men tbgen source
Posted:
15 May 2007, 16:08
by H.G.Muller
If you are interested in the workings of TB generators in general, I have a simple but fast one for Pawn-less endings on my web-site:
http://home.hccnet.nl/h.g.muller/5men13.c
with some descriptions
here.
It cannot do 6 men unless you have an unreasonably big 64-bit system, but it builds a typical 5-men TB in ~240 sec on my 1.3GHz Celeron M. Adapting it to handle Pawns would make it both simpler (you would hardly have to worry about symmetry) and more complicated (you would have to find the DTM of promoted positions from an existing TB).
Re: Looking for Nalimov 5-men tbgen source
Posted:
16 May 2007, 02:11
by Dann Corbit
Here is my output on 2.2 GHz AMD Athlon
- Code: Select all
410020 411000
KBB_KN
0. 0 0 40193555 0 4.984 4.984 0.000
0. 16488 0 40193555 1310904372 38.126 4.984 33.142
1. 21244 437907 69184 0 38.235 5.093 33.142
2. 17139 609625 102148 0 38.376 5.234 33.142
3. 12928 511023 84098 0 38.501 5.359 33.142
4. 11580 420021 65830 0 38.610 5.468 33.142
5. 14292 522067 78481 0 38.735 5.593 33.142
6. 19760 633695 91324 0 38.860 5.718 33.142
7. 28018 1011420 138491 0 39.048 5.906 33.142
8. 43974 1573161 202927 0 39.298 6.156 33.142
9. 62265 2567323 318043 0 39.626 6.484 33.142
10. 98110 4052112 476243 0 40.079 6.937 33.142
11. 140681 6597443 747124 0 40.704 7.562 33.142
12. 210965 9763684 1058627 0 41.532 8.390 33.142
13. 284693 13622055 1446126 0 42.595 9.453 33.142
14. 399918 18804181 1945747 0 43.939 10.797 33.142
15. 562300 22971056 2330980 0 45.548 12.406 33.142
16. 769116 27357987 2729968 0 47.439 14.297 33.142
17. 934463 28350426 2812378 0 49.470 16.328 33.142
18. 842421 20295735 2055593 0 51.298 18.156 33.142
19. 630337 11306075 1270277 0 52.751 19.609 33.142
20. 538938 8133794 948144 0 53.986 20.844 33.142
21. 437436 5464947 629584 0 55.048 21.906 33.142
22. 347383 3865877 441624 0 55.970 22.828 33.142
23. 270176 2838915 328017 0 56.783 23.641 33.142
24. 218373 2152535 251754 0 57.455 24.313 33.142
25. 197301 1860859 221472 0 58.033 24.891 33.142
26. 196877 1829374 218410 0 58.580 25.438 33.142
27. 211266 2000788 241061 0 59.127 25.985 33.142
28. 227531 2172555 258152 0 59.720 26.578 33.142
29. 231384 2252949 265900 0 60.330 27.188 33.142
30. 228903 2292859 266158 0 60.939 27.797 33.142
31. 227364 2233054 257984 0 61.564 28.422 33.142
32. 230553 2278789 257777 0 62.189 29.047 33.142
33. 220597 2290666 257647 0 62.799 29.657 33.142
34. 212820 2215190 246608 0 63.408 30.266 33.142
35. 195737 2120278 235145 0 63.986 30.844 33.142
36. 192307 2012627 220607 0 64.549 31.407 33.142
37. 199594 2037167 220562 0 65.080 31.938 33.142
38. 212780 2161523 231203 0 65.627 32.485 33.142
39. 243478 2384707 251108 0 66.189 33.047 33.142
40. 278943 2742042 286071 0 66.799 33.657 33.142
41. 304942 3080707 321358 0 67.502 34.360 33.142
42. 312444 3242585 340116 0 68.252 35.110 33.142
43. 310188 3212660 335360 0 69.017 35.875 33.142
44. 297251 3087271 321540 0 69.768 36.626 33.142
45. 280541 2901608 299451 0 70.502 37.360 33.142
46. 263536 2705134 276344 0 71.221 38.079 33.142
47. 243629 2577689 261634 0 71.893 38.751 33.142
48. 218826 2383845 241019 0 72.518 39.376 33.142
49. 188894 2048047 207699 0 73.096 39.954 33.142
50. 151543 1656334 168759 0 73.596 40.454 33.142
51. 126050 1320397 133469 0 74.018 40.876 33.142
52. 117183 1189388 119687 0 74.393 41.251 33.142
53. 114785 1137160 113534 0 74.736 41.594 33.142
54. 122978 1117494 112886 0 75.080 41.938 33.142
55. 144843 1152292 115806 0 75.440 42.298 33.142
56. 211461 1416954 141959 0 75.846 42.704 33.142
57. 332282 2058101 203047 0 76.361 43.219 33.142
58. 468435 3074486 300858 0 77.174 44.032 33.142
59. 580855 4213644 405834 0 78.065 44.923 33.142
60. 695064 5224735 499595 0 79.143 46.001 33.142
61. 817064 6326572 596766 0 80.362 47.220 33.142
62. 977336 7523834 700212 0 81.752 48.610 33.142
63. 1212186 8858766 806092 0 83.299 50.157 33.142
64. 1509309 10579075 944487 0 85.049 51.907 33.142
65. 1792981 11939008 1046032 0 87.034 53.892 33.142
66. 1965032 12425970 1068129 0 89.096 55.954 33.142
67. 1992500 11973805 1014735 0 91.174 58.032 33.142
68. 1848015 10609999 891679 0 93.128 59.986 33.142
69. 1642726 8948397 748904 0 94.909 61.767 33.142
70. 1409910 7190366 601680 0 96.487 63.345 33.142
71. 1135315 5331289 444707 0 97.831 64.689 33.142
72. 772689 3326134 276383 0 98.862 65.720 33.142
73. 369030 1409251 116513 0 99.534 66.392 33.142
74. 98149 314052 26218 0 99.862 66.720 33.142
75. 11690 30784 2919 0 99.987 66.845 33.142
76. 3850 8818 966 0 100.034 66.892 33.142
77. 916 2622 300 0 100.065 66.923 33.142
78. 96 613 67 0 100.097 66.955 33.142
79. 0 0 0 0 100.112 66.970 33.142
78958876 marked win.wtm
384348377 labeled potential win(n).btm
1310904372 checked
314204330 2377132303
Press any key to continue . . .
Re: Looking for Nalimov 5-men tbgen source
Posted:
21 May 2007, 12:45
by H.G.Muller
Yes, that is how it is supposed to look. The meaning of the columns is:
1 - DTM (0 = checkmated)
2 - Number of positions with this DTM for black to move
3 - Number of positions with this DTM+1 for white to move
4 - Number of btm positions that got their non-losing-move count reduced or set
5 - Number of moves tested for being losing from btm posiitons
6 - total cumulative execution time in sec
7 - cumulative time spent in finding 2 generations of pre-decessor positions of the current DTM
8 - cumulative time spent in verifying potentially lost btm positions
So for the KBBKN end-game the generation of the TB took 33 sec for counting the number of legal moves in all BTM positions, and 67 sec to determine all positions won for white and the DTM of all positions lost to black. Largest DTM was 78, and there were 96 (btm) positons that had this DTM.
Note the number of positions are those folded by symmetry, i.e. not corrected for the number of actual positions that correspond to the symmetry-reduced TB elements, and furthermore contain all subset end-games as well (so KBBK, KBKN, KBK, KKN and KK in this case). This makes them useless for comparing to other tablebase statistics, they only serve to see how your algorithm is progressing.
Meaningful statistics, corrected for symmetry and only including positons with the nominal piece makeup are written in a file 'report.txt' (but still have to be multiplied by 4).