Gull 3 Linux+Syzygy and Fathom released.

[stuwph]

Hey! Good day folks! Is there any Gull non-bmi2 version? I would love to have it. Gull is too exciting to watch in engine matches.

Regards,
Angelo

here now is is also a popcnt version, perhaps it works on your pc
http://1drv.ms/1Qb2V48

[Stonewall]

thanks brother ! I got core i-5 probably works for now. again, thanks a lot! Have a nice day to you and your family!

Regards,
Angelo

[basil00]

here now is is also a popcnt version, perhaps it works on your pc
http://1drv.ms/1Qb2V48

Thanks for making this available.

Please post test results if anyone has them. Especially:

• Gull 3+syzygy windows versus the official Gull 3 windows
• Any test on TCEC-like hardware (high number of cores).

Also note to anyone testing that Gull Syzygy will always report TBHits as 0, even if there are TB hits. This is a bug on the TODO list.

[Dan Cooper]

here now is is also a popcnt version, perhaps it works on your pc
http://1drv.ms/1Qb2V48

Thanks for making this available.

Please post test results if anyone has them. Especially:

• Gull 3+syzygy windows versus the official Gull 3 windows
• Any test on TCEC-like hardware (high number of cores).

Also note to anyone testing that Gull Syzygy will always report TBHits as 0, even if there are TB hits. This is a bug on the TODO list.

Do you plan on submitting any dev versions to TCEC?

[arunsoorya1309]

Basil, people are saying you don't want to participate this GULL version in TCEC, is it true, any specific reason if any.

Would be nice if you contact Anton and submit this version. Since its only stage 1 and gull surely will get selected for stage 2 , it may be good to have a new version for stage 2

[Damir]

Hey! Good day folks! Is there any Gull non-bmi2 version? I would love to have it. Gull is too exciting to watch in engine matches.

Regards,
Angelo

here now is is also a popcnt version, perhaps it works on your pc
http://1drv.ms/1Qb2V48

Thanks for the new version Uwe. However it seems is not using syzygy tbs.
I have tried and test it offline, and there are no tbs hits. It also played this very strange game, where Stockfish saw it had clear advantage and was winning, while Gull only saw the game as draw. Perhaps bug in the evaluation? Here it is:

[Event "Zagor Almanah, Blitz 1m+1s"]
[Site "VozackaTeorija"]
[Date "2016.05.03"]
[Round "1"]
[White "Stockfish 160316 64 POPCNT"]
[Black "Gull 3 x64"]
[Result "1-0"]
[ECO "C54"]
[Annotator "0.29;0.26"]
[PlyCount "125"]
[TimeControl "60+1"]

{Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz 3956 MHz W=27.1 plies;
9354kN/s; Crusher v70.ctg B=16.4 plies; 2649kN/s; Crusher v70.ctg} 1. e4 {B 0}
e5 {B 0} 2. Nf3 {B 0} Nc6 {B 0} 3. Bc4 {B 0} Nf6 {B 0} 4. d3 {B 0} Bc5 {B 0} 5.
c3 {B 0} O-O {B 0} 6. O-O {B 0} d6 {B 0} 7. h3 {B 0} h6 {B 0 Both last book
move} 8. b4 {0.29/17 1} Bb6 {0.26/15 2} 9. a4 {0.15/22 11} a5 {0.33/17 2} 10.
b5 {0.13/21 1} Ne7 {0.33/16 0} 11. Re1 {0.19/21 5 (Nbd2)} Ng6 {0.29/18 3} 12.
d4 {0.04/21 3 (Nbd2)} exd4 {-0.16/16 1} 13. cxd4 {0.02/22 2} Nxe4 {-0.20/16 1}
14. Rxe4 {-0.10/17 0} d5 {-0.20/17 1} 15. Bxd5 {0.00/23 1} Qxd5 {-0.20/16 0}
16. Nc3 {0.00/24 0} Qc4 {-0.20/16 0} 17. Ra3 {0.00/25 1} f6 {-0.15/17 1} 18.
Ne2 {0.00/24 6} Rd8 {-0.25/18 5 (Bf5)} 19. Rc3 {-0.02/23 2} Qf7 {-0.25/17 2}
20. Ba3 {0.00/21 0 (Ng3)} f5 {-0.03/18 3 (Be6)} 21. Ree3 {-0.10/22 2} f4 {
-0.03/18 1} 22. Ne5 {0.00/22 2} Qf6 {-0.03/17 0} 23. Red3 {0.00/21 0 (Nxg6)}
Bf5 {-0.41/16 2} 24. Rd2 {-0.07/22 2 (Nxg6)} Nxe5 {-0.42/17 3} 25. dxe5 {
-0.04/22 2} Rxd2 {-0.42/16 0} 26. Qxd2 {-0.14/21 0} Qxe5 {-0.42/16 0} 27. Nxf4
{-0.24/23 3} Re8 {-0.42/16 0} 28. Rc1 {-0.14/21 0} Be4 {-0.22/16 3} 29. Bb2 {
-0.05/24 4} Qd6 {-0.15/18 5} 30. Rd1 {-0.06/24 2} Rd8 {-0.16/19 4 (Qxd2)} 31.
Qxd6 {-0.08/28 2} Rxd6 {-0.13/20 2} 32. Rxd6 {-0.11/27 0} cxd6 {-0.13/19 0} 33.
Ne6 {-0.07/31 2} Bc2 {-0.13/19 0 (Kf7)} 34. Bxg7 {-0.07/27 1} Bxa4 {-0.11/20 6}
35. Bd4 {-0.07/29 1} Bc5 {-0.11/19 0} 36. Nxc5 {-0.07/31 2} dxc5 {-0.11/17 0}
37. Bxc5 {-0.07/29 1} Bxb5 {-0.11/19 3} 38. f4 {-0.07/32 1 (g4)} Bc6 {-0.10/19
7 (Kf7)} 39. g4 {0.00/26 1} b5 {-0.10/17 0 (Kf7)} 40. Kf2 {0.50/23 1} b4 {
-0.10/16 4} 41. f5 {0.59/24 0 (Ke3)} b3 {-0.08/13 2 (Bd5)} 42. Ba3 {0.60/31 1}
h5 {0.00/16 2 (Kf7)} 43. Kg3 {1.39/26 1} Be4 {0.07/17 2} 44. Kf4 {1.57/30 0
(Kh4)} Bg2 {0.14/17 4} 45. Kg5 {4.62/38 11} Kg7 {0.15/16 1} 46. Bb2+ {5.22/24
1 (Kxh5)} Kf8 {0.15/18 2 (Kg8)} 47. Kxh5 {5.57/23 1} Bxh3 {0.15/16 0} 48. Kh4 {
7.20/28 1 (Ba3+)} Bg2 {0.73/17 12} 49. g5 {12.64/29 3} Be4 {0.73/16 0} 50. Kg4
{13.48/19 0} Kf7 {0.73/16 0 (Bc6)} 51. g6+ {39.77/25 2} Kg8 {1.10/17 11} 52.
Kg5 {47.11/32 1} a4 {1.42/17 12 (Bc6)} 53. f6 {52.15/35 1} Bd5 {1.42/17 0} 54.
Ba3 {67.41/35 1 (Kf5)} Bc4 {1.54/16 2 (Be6)} 55. Kf4 {110.59/34 2} b2 {1.70/17
2} 56. Bxb2 {#22/34 2} Bb3 {2.18/16 1 (Kf8)} 57. Ba3 {#13/38 1 (Ke5)} Ba2 {
4.72/14 1} 58. Ke5 {#12/38 0 (Bc5)} Bb3 {7.18/9 1 (Bc4)} 59. Kd6 {#11/39 1} Kh8
{13.57/10 1 (Bf7)} 60. Ke7 {#8/32 1} Bd1 {#9/10 1 (Bf7)} 61. Bb2 {#3/40 1 (f7)}
a3 {#2/19 1} 62. f7+ {#2/50 0 (Bd4)} axb2 {#2/6 0} 63. f8=Q# {1/0 0} 1-0

[velmarin]

Thanks for the new version Uwe. However it seems is not using syzygy tbs.

It is not in the code "output TBhits".
I already got it, and works well, but the "Fake PV" that requires more work.
A workaround of joke for this great engine

Code: Select all

printf("info depth 1 seldepth 1 score cp %d nodes 1 nps 0 pv %s\n", best_score, str);   // Fake PV

[basil00]

Basil, people are saying you don't want to participate this GULL version in TCEC, is it true, any specific reason if any.

Not true -- they are welcome to use it if they want. I am not the official Gull dev so it is not my decision.

It is not in the code "output TBhits".

That's true, it is not been implemented. Gull's codebase is opaque so it is not so simple to add even basic features like this.

[velmarin]

It is not in the code "output TBhits".

That's true, it is not been implemented. Gull's codebase is opaque so it is not so simple to add even basic features like this.

I've added github, I think...
I'm not a connoisseur of Github.
https://github.com/Velmarin/Gull/tree/patch-1

It works on Windows for me.

[basil00]

I've added github, I think...
I'm not a connoisseur of Github.
https://github.com/Velmarin/Gull/tree/patch-1

This will not be accurate since the tb_hits counter is not updated in a thread-safe way. This can be fixed by using atomic increment, but that could slow the entire engine down. Better is to have a local tb_hits per thread (a.k.a. process), and then tally all the tb_hits at the end.