Testing new compiles of Stockfish.

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Harvey Williamson, bob

Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 7:09 pm

Re: Testing new compiles of Stockfish.

Post by Tomcass » Thu Mar 22, 2012 7:12 pm

Gary, this is the answer I have received from Quocvuong:

Source is being used in Stockfish 120321: I get from Github Wednesday, ‎March ‎21, ‎2012, Link download (Marco Costalba):

http://www.mediafire.com/?w1km0hkatxyuzkb

Regards,

Tom.

gladius
Posts: 535
Joined: Tue Dec 12, 2006 9:10 am

Re: Testing new compiles of Stockfish.

Post by gladius » Thu Mar 22, 2012 8:15 pm

Great, thanks Tom. Just wondering if the undefended pieces change was in there, and v1 of it was.

Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 7:09 pm

Re: Testing new compiles of Stockfish.

Post by Tomcass » Tue Mar 27, 2012 10:13 am

StockfishQUO20120323 119 GAMES

I7 980 3.33 Ghz.
Critter, Houdini, Deep Rybka 6 real cores - NO HT
Stockfish 6 real cores + 2 HT
GUI: Fritz 12
Book: Perfect 2012
Time control: 2 minutes +2 sec/ game.
Ponder: Off
No tablebases

Stockfish 120323MPx8 - Critter 1.4a 64-bit SSE4(x6) 19.5 - 20.5 +7/=25/-8 48.75%
Stockfish 120323MPx8 - Deep Rybka 4.1 SSE42 x64 (x6) 21.0 - 19.0 +13/=16/-11 52.50%
Stockfish 120323MPx8 - Houdini 2.0c Pro x64(x6) 18.5 - 20.5 +11/=15/-13 47.44%

http://www.mediafire.com/?il63uf6fqg3vgrv

STOCKFISH QUO 120323 120 GAMES

i7 9750 3.33 Ghz.
4 real cores Critter DR and Houdini
4 real cores + 2 HT Stockfish
GUI: Fritz 12
Book: Masterbook 3.1
Time control: 2 minutes + 2 sec/game.
Ponder: Off.
No tablebases

Stockfish 120323MP6 - Critter 1.4a 64-bit SSE4x4 18.5 - 21.5 +8/=21/-11 46.25%
Stockfish 120323MP6 - Deep Rybka 4.1 SSE42 x64(x4) 19.0 - 21.0 +10/=18/-12 47.50%
Stockfish 120323MP6 - Houdini 2.0c Pro x64(x4) 19.5 - 20.5 +8/=23/-9 48.75%

http://www.mediafire.com/?7zyp4nuuy0dwh6c


GLOBAL TEST FOR STOCKFISH 120323QUO AFTER 239 GAMES:

Stockfish 120323MP - Critter 1.4a 64-bit SSE4 38.0 -42.0 +15/=26/-17 47.50%
Stockfish 120323MP - Deep Rybka 4.1 SSE42 x64 40.0 - 40.0 +23/=34/-23 50.00%
Stockfish 120323MP - Houdini 2.0c Pro x64 38.0 – 41.0 +19/=38/-22 48.10%

Average Result: 48.54%

This is a strong version. One of the best results of Stockfish ever against Houdini 2.0c Pro in my tests.

Regards from Barcelona,

Tom.

Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 7:09 pm

Re: Testing new compiles of Stockfish.

Post by Tomcass » Tue Apr 03, 2012 8:12 am

TEST STOCKFISH 120331 PETERPAN

i7 9750 3.33 Ghz.
4 real cores Critter DR and Houdini
4 real cores + 2 HT Stockfish INTEL
GUI: Fritz 12
Book: Masterbook 3.1
Time control: 2 minutes + 2 sec/game.
Ponder: Off.
No tablebases

Stockfish 120331P6Int - Critter 1.4a 64-bit SSE4x4 18.5 - 21.5 +5/=27/-8 46.25%
Stockfish 120331P6Int - Deep Rybka 4.1 SSE42 x64(x4) 18.0 - 22.0 +3/=30/-7 45.00%
Stockfish 120331P6Int - Houdini 2.0c Pro x64(x4) 13.0 - 27.0 +4/=18/-18 32.50%

http://www.mediafire.com/?19da71cenelaeqr

I7 980 3.33 Ghz.
Critter, Houdini, Deep Rybka 6 real cores - NO HT
Stockfish 6 real cores + 2 HT AMD
GUI: Fritz 12
Book: Perfect 2012
Time control: 2 minutes +2 sec/ game.
Ponder: Off
No tablebases

Stockfish 120331P8Amd - Critter 1.4a 64-bit SSE4(x6) 21.0 - 19.0 +11/=20/-9 52.50%
Stockfish 120331P8Amd - Deep Rybka 4.1 SSE42 x64 (x6) 21.0 - 19.0 +11/=20/-9 52.50%
Stockfish 120331P8Amd - Houdini 2.0c Pro x64(x6) 18.0 - 22.0 +7/=22/-11 45.00%

http://www.mediafire.com/?nwfcsilebo9scbf


GLOBAL TEST FOR STOCKFISH20120331 PETERPAN AFTER 240 GAMES.

Stockfish 120331PInt - Critter 1.4a 64-bit SSE4 39.5 -40.5 +16/=47/-17 49.37%
Stockfish 120331PInt - Deep Rybka 4.1 SSE42 x64 39.0 - 41.0 +14/=50/-16 48.75%
Stockfish 120331PInt - Houdini 2.0c Pro x64 31.0 - 49.0 +11/=40/-29 38.75%

Average result: 45.63%


The Intel version seems weak, although its speed is not bad. However the AMD version has shown a relevant strength in this test. (Please note that my computers are not AMD but INTEL).

Regards from Barcelona.

Tom

Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 7:09 pm

A very strong AMD compile.

Post by Tomcass » Fri Apr 06, 2012 9:33 am

I was surprised by the performance of Stockfish 120331 PeterPan (Izak) with 6 real cores (+ 2 HT) and I decided to repeat the test. Here you have the results:

I7 980 3.33 Ghz.
Critter, Houdini, Deep Rybka 6 real cores - NO HT
Stockfish120331 AMD 6 real cores + 2 HT
GUI: Fritz 12
Book: Perfect 2012
Time control: 2 minutes +2 sec/ game.
Ponder: Off
No tablebases

Stockfish 120331P8Amd - Critter 1.4a 64-bit SSE4(x6) 20.5 - 19.5 +9/=23/-8 51.25%
Stockfish 120331P8Amd - Deep Rybka 4.1 SSE42 x64 (x6) 23.5 - 16.5 +11/=25/-4 58.75%
Stockfish 120331P8Amd - Houdini 2.0c Pro x64(x6) 18.5 - 21.5 +7/=23/-10 46.25%

http://www.mediafire.com/?sm3m8vj19cpfnyr

Previous match under the same conditions (AMD compile):

Stockfish 120331P8Amd - Critter 1.4a 64-bit SSE4(x6) 21.0 - 19.0 +11/=20/-9 52.50%
Stockfish 120331P8Amd - Deep Rybka 4.1 SSE42 x64 (x6) 21.0 - 19.0 +11/=20/-9 52.50%
Stockfish 120331P8Amd - Houdini 2.0c Pro x64(x6) 18.0 - 22.0 +7/=22/-11 45.00%

http://www.mediafire.com/?nwfcsilebo9scbf

If we add the previous match I did under these conditions, we have that

AFTER 240 GAMES THE GLOBAL RESULT IS:

Stockfish 120331P8Amd - Critter 1.4a 64-bit SSE4(x6) 41.5 – 38.5 +20/=43/-17 51.87%
Stockfish 120331P8Amd - Deep Rybka 4.1 SSE42 x64 (x6) 44.5 – 35.5 +22/=45/-13 55.62%
Stockfish 120331P8Amd - Houdini 2.0c Pro x64(x6) 36.5 – 43.5 +14/=45/-22 45.62%

AVERAGE GLOBAL RESULT: 51.04%


I know that 240 games are not enough to prove anything, but this is the strongest version of Stockfish I have tested till now. Please note that:

- The INTEL version seems to be much weaker (although my computers are INTEL based!)
- I have added a bit of ‘doping’ with two HT cores. Although Houdini and Critter do not benefit from HT, both Stockfish and Ivanhoe play slightly better with them.
- I would appreciate if testers with very fast machines can work on this AMD version to confirm these results or not. Thanks in advance.

Regards from Barcelona.

Tom.

WuShock
Posts: 162
Joined: Thu Jul 19, 2007 1:13 am

Re: A very strong AMD compile.

Post by WuShock » Mon Apr 16, 2012 3:23 pm

Tested Stockfish 120331P --- AMD compile

i7 940 / 3.75 GHz / ponder off / no HT , no LP / 2 threads each
F11 GUI / klo_50_eco openings / 512 MB / all x64
SF + Critter + Rybka = no egtb
Houdini used 3-4-5 Nalimov , though didn't mean to..........

Time Control : 4 mins + 2 secs


SF 120331P...........55.5............+32,-21,=47
Rybka 4.1..............44.5

SF120331P............52.5............+25,-20,=55
Critter 1.4..............47.5

Houdini 2.0c...........51.5............+28,-25,=47
SF120331P............48.5

---------------------------------------------------------------

Couple games where SF wins both sides of the Sicilian.........

[Date "2012.04.15"]
[White "Stockfish 120331P"]
[Black "Houdini 2.0c x64"]
[Result "1-0"]
[ECO "B21"]

1. e4 c5 2. d3 Nc6 3. f4 d6 4. Nc3 g6 {Both last book move} 5.
Nf3 {0.04/20 6} Nf6 {-0.29/19 12} 6. Be2 {(Nd5) 0.08/20 6} Bg7 {-0.27/19 11} 7.
O-O {0.00/20 10} O-O {-0.25/18 11} 8. Kh1 {(Qe1) 0.04/22 18} Rb8 {(Bd7) -0.31/
17 7} 9. a3 {(a4) 0.20/19 10} b5 {-0.30/16 5} 10. Bd2 {(Rb1) 0.20/19 6} Qb6 {
(Bd7) -0.27/17 11} 11. Be1 {(Rb1) 0.16/18 10} Bb7 {(Be6) -0.37/17 8} 12. Bh4 {
(f5) 0.00/19 6} Nd4 {(e6) -0.38/18 10} 13. Nxd4 {0.08/19 6} cxd4 {-0.38/16 0}
14. Nb1 {(Bf2) 0.12/21 14} Rfc8 {(Rbe8) -0.20/17 19} 15. c3 {(Nd2) 0.28/20 7}
Rc7 {(dxc3) -0.12/16 7} 16. Bf2 {0.36/20 7} Nd7 {-0.10/19 5} 17. Nd2 {0.00/21
14} Nc5 {-0.07/20 21} 18. cxd4 {0.04/22 8} Na4 {-0.11/22 4} 19. Nf3 {(d5) 0.00/
24 12} Nxb2 {-0.02/20 9} 20. Qd2 {0.00/24 6} Na4 {-0.02/19 6} 21. d5 {0.00/23 8
} Qa6 {0.07/20 23} 22. Nd4 {(f5) 0.00/23 6} Nc3 {-0.36/18 4} 23. Bf3 {0.16/23
18} Qa4 {-0.26/17 8} 24. Qe3 {0.00/23 5} Re8 {-0.20/17 18} 25. f5 {(Bg4) 0.20/
22 8} Bf6 {(Bxd4) -0.09/16 4} 26. fxg6 {(Rfc1) 0.56/19 6} hxg6 {1.72/17 19} 27.
Ne6 {1.71/20 7} Rec8 {(Nxd5) 2.10/16 6} 28. Nxc7 {3.11/24 11} Rxc7 {2.33/18 4}
29. h4 {(Qxa7) 3.57/22 18} Bc8 {1.39/15 4} 30. Rfc1 {(h5) 4.12/21 10} Bg7 {1.
85/16 6} 31. h5 {(d4) 4.58/21 6} gxh5 {(Bb7) 2.27/17 7} 32. Bxh5 {(d4) 4.56/12
0} Bd7 {(Qa5) 2.69/15 3} 33. Qf3 {(d4) 5.41/19 5} f5 {4.19/16 12} 34. Be3 {
(Bg6) 5.77/19 6} b4 {4.31/14 3} 35. axb4 {6.62/18 2} Qb5 {5.24/15 9} 36. Bxa7 {
7.27/18 5} Qxb4 {6.19/15 14} 37. Be3 {7.27/12 0} Rc8 {6.94/16 10} 38. Qg3 {
(Bg6) 9.67/21 6} Be8 {(f4) 7.27/15 2} 39. Bxe8 {(Bh6) 12.04/22 3} Rxe8 {10.64/
17 7} 40. Qg6 {12.72/23 3} Rf8 {10.84/18 5} 41. Ra7 {14.42/24 3} Nxe4 {(Nxd5)
11.72/18 4} 42. dxe4 {(Rxe7) 95.77/23 14} Qb2 {24.21/18 4} 43. e5 {(Rc8) #12/
18 3} Qb4 {(Qxe5) 31.89/16 2} 44. Bg5 {(Qe6+) #6/26 3} Qb2 {(Qe4) #5/20 0} 45.
Rxe7 {#5/54 12} Qxe5 {#4/18 0} 46. Rxe5 {#4/12 0} dxe5 {#3/16 0} 47. Rc7 {(Bh6)
#3/12 0} Rf7 {#2/14 0} 48. Qxf7+ {#2/256 0} Kh7 {#1/12 0} 49. Qxg7# {#1/256 0}
1-0


[Date "2012.04.15"]
[White "Houdini 2.0c x64"]
[Black "Stockfish 120331P"]
[Result "0-1"]
[ECO "B21"]


1. e4 c5 2. d3 Nc6 3. f4 d6 4. Nc3 g6 {Both last
book move} 5. g4 {-0.16/19 8} Bg7 {(Nf6) -0.16/19 10} 6. g5 {(h3) -0.08/19 19}
h6 {-0.32/21 14} 7. Bg2 {(Nf3) -0.07/19 5} hxg5 {-0.56/21 6} 8. fxg5 {-0.07/17
0} Be5 {(Rb8) -0.72/22 6} 9. Nf3 {(Nce2) -0.24/20 5} Bg4 {-0.80/23 7} 10. Qd2 {
(Ne2) -0.28/21 7} Bxf3 {-0.76/20 9} 11. Bxf3 {-0.27/20 0} Nd4 {-0.88/22 9} 12.
Qf2 {-0.32/20 14} Qa5 {(e6) -0.60/22 23} 13. Kd1 {(Bd1) -0.28/19 5} e6 {(c4)
-0.88/18 10} 14. h4 {(Bd2) -0.25/18 6} Ne7 {-1.09/19 9} 15. Bg4 {-0.27/18 5}
Nec6 {(Kd7) -0.96/20 8} 16. Bd2 {-0.18/18 8} O-O-O {(Qa6) -1.17/21 7} 17. Ne2 {
(Qxf7) -0.34/17 10} Qb5 {-1.09/21 20} 18. Bc3 {-0.32/18 6} Rdf8 {(Qb6) -1.21/
21 5} 19. a4 {-0.42/19 5} Qb6 {-0.84/20 15} 20. a5 {-0.41/19 4} Qb5 {-0.96/21
16} 21. a6 {-0.49/19 25} b6 {-1.01/21 6} 22. Rf1 {(Kc1) -0.49/18 4} Kd7 {(Nxe2)
-1.37/20 8} 23. Kc1 {(Ra2) -0.48/18 5} f6 {-1.57/20 9} 24. Nxd4 {(gxf6) -0.69/
18 14} cxd4 {(Nxd4) -1.57/22 7} 25. Be1 {-0.86/20 20} fxg5 {-1.73/24 5} 26.
Qxf8 {-1.12/19 5} Rxf8 {-1.73/12 0} 27. Rxf8 {-0.96/18 0} Ke7 {-2.34/24 8} 28.
Rg8 {(Ra8) -0.83/20 4} gxh4 {(Kf7) -2.34/23 6} 29. Bxh4+ {-0.83/20 4} Kf7 {-2.
34/13 0} 30. Ra8 {-0.83/21 5} Bf6 {-2.10/26 17} 31. Bxf6 {-0.83/20 0} Kxf6 {-2.
10/12 0} 32. Kb1 {(Rc8) -1.04/22 8} Qb4 {-2.78/24 6} 33. Rf8+ {(Bd1) -1.04/22 4
} Ke5 {(Ke7) -3.27/25 5} 34. Rf1 {-1.78/23 21} Qd2 {-3.51/25 4} 35. Rc1 {(Bd1)
-1.79/22 17} Qg2 {-3.71/23 5} 36. Bd1 {-1.87/20 5} Kf4 {(b5) -3.67/24 8} 37. c3
{(Ra3) -1.75/18 3} g5 {-3.55/22 15} 38. Ra3 {-2.16/19 9} Qd2 {-4.72/22 5} 39.
Ba4 {-2.46/20 15} Qxd3+ {-4.60/21 2} 40. Ka1 {-2.52/18 3} Qe2 {(Na5) -5.41/22 5
} 41. Bxc6 {-3.26/19 11} d3 {-5.41/13 0} 42. Ba4 {(Rb1) -2.93/19 3} b5 {(d2)
-5.81/22 2} 43. Bb3 {-3.70/19 3} g4 {(d2) -5.81/22 4} 44. Kb1 {-3.95/19 4} d2 {
-5.81/24 3} 45. Rd1 {-3.95/18 0} g3 {-5.89/26 4} 46. Ra1 {-4.17/20 3} g2 {
(Qxd1+) -6.02/26 11} 47. Ka2 {-3.79/19 5} b4 {-6.42/21 2} 48. cxb4 {-3.81/20 4}
Qxa6+ {-6.42/12 0} 49. Kb1 {-3.80/19 0} Qd3+ {(Qc6) -7.03/20 4} 50. Ka2 {-2.92/
20 3} d5 {(Qb5) -7.95/20 3} 51. e5 {-4.83/19 8} Kxe5 {-9.33/20 2} 52. Rg1 {
(Ba4) -5.53/19 8} Kf4 {(Qa6+) -10.47/24 9} 53. Rad1 {(Ba4) -6.46/18 6} Kf3 {
-11.15/21 2} 54. Ba4 {-5.75/16 0} Kf2 {-11.31/22 2} 55. Bd7 {-7.35/17 5} e5 {
-11.77/21 3} 56. b3 {(b5) -7.57/18 2} e4 {-13.17/20 2} 57. Ka3 {-9.86/18 4} Kg3
{(Qe3) -14.18/22 2} 58. Bf5 {(Be8) -9.49/17 4} Qd4 {-21.05/19 3} 59. Bg4 {-12.
21/16 3} Kh2 {-25.37/20 3} 60. Rxg2+ {-13.61/17 3} Kxg2 {-25.37/12 0} 61. Bh5 {
-15.15/18 3} e3 {-44.28/18 2} 62. b5 {(Ka4) -16.14/18 2} Kf2 {-101.77/22 6} 63.
b6 {(Rh1) -19.82/17 2} axb6 {(Qxb6) -101.77/18 3} 64. Rh1 {-#8/18 2} e2 {-#8/
21 1} 65. Bxe2 {-#7/24 0} Kxe2 {-#7/12 0} 66. Ka2 {(b4) -#6/22 0} d1=Q {-#6/36
2} 67. Rxd1 {-#5/20 0} Kxd1 {-#5/88 2} 68. Kb1 {-#4/18 0} Qd2 {(Kd2) -#4/107 2}
69. b4 {-#3/16 0} Qxb4+ {-#3/256 0} 70. Ka2 {-#2/1 0} Kc1 {(Kc2) -#2/256 0} 71.
Ka1 {-#1/1 0} Qb1# {-#1/256 0} 0-1

Tomcass
Posts: 786
Joined: Sun Apr 16, 2006 7:09 pm

Re: A very strong AMD compile.

Post by Tomcass » Mon Apr 16, 2012 3:31 pm

Thanks for posting this very good test, Tom. It confirms my previous results. This is one of the best compiles of Stockfish ever. Thanks to Marco & Team for developing Stockfish and also to Izak for this strong compile.

Regards,

Tom Cass

ernest
Posts: 1746
Joined: Wed Mar 08, 2006 7:30 pm

Re: A very strong AMD compile.

Post by ernest » Mon Apr 16, 2012 5:56 pm

Tomcass wrote:Thanks for posting this very good test, Tom.
On my Core 2 duo, I could not test the popcount versions.
So all I could do, is run the normal mingw compile Stockfish 120331M x64

And that one seems only on par (no better) with my "official" stockfish-222-64-ja-intel... :(

Post Reply