Komodo 11.3

[Hurnavich]

Just a quick question Contempt changed to 16 default should this be 0 against = or higher opposition I notice on my system seems slower in k/ns by about 600 thanks in advance.

[Hurnavich]

A quick eg of speed diff on my AMD depth appears a little better


[Event "K 11.3 vs K 11.2.2, Blitz 5m+2s"]
[Site "Hurnavich-PC"]
[Date "2018.03.04"]
[Round "1"]
[White "Komodo 11.3 64-bit LTC"]
[Black "Komodo 11.2.2 64-bit LTC"]
[Result "1/2-1/2"]
[ECO "A00"]
[Annotator "-0.07;-0.01"]
[PlyCount "103"]
[TimeControl "300+2"]

{AMD Phenom(tm) II X6 1100T Processor 3625 MHz W=26.4 plies; 6,997kN/s; 615,
771 TBAs B=26.0 plies; 7,923kN/s; 988,953 TBAs} 1. b4 d5 2. Bb2 Bg4 {Both
last book move} 3. h3 {-0.07/23 17} Bf5 {-0.01/22 10 (Bh5)} 4. e3 {0.07/23 8}
e6 {0.00/24 16 (Nf6)} 5. c4 {-0.07/23 10} Nf6 {0.05/22 5 (h6)} 6. Nf3 {0.00/23
9} Bxb4 {0.00/25 20 (h6)} 7. cxd5 {0.10/25 9} Qxd5 {0.15/26 11} 8. Qa4+ {0.23/
26 9} Nc6 {0.18/25 4} 9. Bb5 {0.19/27 8} Bd6 {0.14/27 8} 10. Bxf6 {0.22/26 9}
gxf6 {0.13/26 10} 11. Nc3 {0.16/26 8} Qc5 {0.15/26 6} 12. Nd4 {0.22/26 10} O-O
{0.16/25 6} 13. Bxc6 {0.23/27 22 (Nxf5)} bxc6 {0.17/24 5} 14. Nxf5 {0.21/27 19}
Qxf5 {0.15/26 9} 15. Ke2 {0.22/26 16} Rab8 {0.26/24 9 (Be5)} 16. Rab1 {0.19/25
17} Be7 {0.22/24 7} 17. Rxb8 {0.21/25 13} Rxb8 {0.19/25 5} 18. Qxc6 {0.18/26 10
} Qa5 {0.23/26 12 (Bd6)} 19. Rc1 {0.24/26 11} Kg7 {0.21/25 13} 20. Ne4 {0.19/
27 11 (a4)} f5 {0.21/25 7} 21. Qc3+ {0.25/28 16} Qxc3 {0.23/25 3} 22. Nxc3 {0.
20/27 10} Rb6 {0.30/25 6} 23. Nd1 {0.27/27 7} Bd6 {0.25/24 4} 24. Rc4 {0.15/27
8} Kf6 {0.40/25 25 (h5)} 25. Rh4 {0.38/26 6 (Kd3)} Kg7 {0.34/27 19 (Bb4)} 26.
Ra4 {0.26/29 9} a6 {0.36/27 3} 27. f3 {0.34/27 5} h5 {0.29/27 4 (Kg6)} 28. Nf2
{0.36/27 7} Kf6 {0.33/26 4 (f6)} 29. Nd3 {0.36/28 6} Rb1 {0.29/27 15 (Kg6)} 30.
f4 {0.26/26 7} Rb6 {0.26/26 2} 31. Nc1 {0.20/28 5} Ke7 {0.25/28 3} 32. Nb3 {0.
19/29 17} Kd7 {0.27/29 12} 33. Kd3 {0.23/26 3} Bb4 {0.27/30 7 (Kc6)} 34. Nd4 {
0.22/28 4 (Kc4)} Bc5 {0.19/26 3 (Bd6)} 35. Nf3 {0.22/31 3} f6 {0.24/25 5 (Bd6)}
36. Kc4 {0.26/27 11} Rc6 {0.05/26 3} 37. Kd3 {0.12/27 4 (d4)} Rb6 {0.26/27 5}
38. Kc4 {0.11/26 15 (h4)} Rc6 {0.14/28 4} 39. Kb3 {0.21/26 3} Rb6+ {0.11/27 2
(Bd6)} 40. Kc3 {0.11/27 4} a5 {0.12/27 6} 41. Kc4 {0.00/27 5} Rc6 {0.21/27 8}
42. Kd3 {0.07/27 3} Ra6 {0.22/25 3} 43. Rc4 {0.12/26 2 (Ke2)} Bb4 {0.13/26 5}
44. e4 {0.14/26 2} Rd6+ {0.14/26 4} 45. Ke3 {0.00/28 5 (Ke2)} fxe4 {0.06/26 3}
46. Rxe4 {0.00/26 2} Rd5 {0.08/28 7} 47. Rc4 {0.00/28 2} Rb5 {0.00/29 8} 48. g4
{0.00/26 3 (a4)} Bc5+ {0.00/27 7 (Bd6)} 49. d4 {0.00/24 2 (Ke2)} Bd6 {0.00/28 3
} 50. Ra4 {0.00/27 2 (Rc2)} Ke7 {0.00/29 4} 51. Nh4 {0.00/28 2 (g5)} c5 {0.00/
26 4 (hxg4)} 52. f5 {0.00/25 2 Draw accepted} 1/2-1/2

[lkaufman]

Just a quick question Contempt changed to 16 default should this be 0 against = or higher opposition I notice on my system seems slower in k/ns by about 600 thanks in advance.

Yes, use zero for Contempt against equal or stronger opposition. The default is probably about right for a 200 elo difference or so.
k/ns should be a bit lower with 11.3 due to a change in how illegal king moves are handled; it's not a real reduction but just a more honest count. Reported depth should be somewhat higher, which is due to several changes and accounts for the elo gain more or less.

[Hurnavich]

Thanks Larry clears that up i notice depth is a little Higher.

[Event "K 11.3 vs K 11.2.2, Blitz 5m+2s"]
[Site "Hurnavich-PC"]
[Date "2018.03.04"]
[Round "1"]
[White "Komodo 11.3 64-bit LTC"]
[Black "Komodo 11.2.2 64-bit LTC"]
[Result "1/2-1/2"]
[ECO "C01"]
[Annotator "0.21;0.09"]
[PlyCount "165"]
[TimeControl "300+2"]

{AMD Phenom(tm) II X6 1100T Processor 3625 MHz W=27.1 plies; 6,643kN/s; 2,827,
139 TBAs B=26.5 plies; 7,542kN/s; 3,484,135 TBAs} 1. e4 {0.21/24 21 Both last
book move} e6 {0.09/23 12} 2. d4 {0.17/25 13 (Nf3)} d5 {0.20/23 5} 3. exd5 {0.
19/26 22 (Nd2)} exd5 {0.09/23 5} 4. Nf3 {0.15/22 8} Nf6 {0.09/22 7} 5. Bd3 {0.
19/24 16 (Nc3)} Bd6 {0.06/25 32} 6. h3 {0.11/24 24 (0-0)} O-O {0.08/24 13} 7.
O-O {0.09/24 12} h6 {0.05/25 25} 8. Re1 {0.18/24 13 (Na3)} Re8 {0.14/24 10 (a6)
} 9. Rxe8+ {0.20/25 8} Qxe8 {0.20/24 4} 10. Nc3 {0.21/25 10} Bd7 {0.14/25 17}
11. Be3 {0.23/25 8} Nc6 {0.21/26 14} 12. Nb5 {0.33/26 7} Nb4 {0.35/25 7} 13.
Nxd6 {0.36/27 7} cxd6 {0.33/25 5} 14. Be2 {0.35/27 14} Bf5 {0.33/25 11 (Ba4)}
15. Ne1 {0.39/27 13} Nc6 {0.33/24 3} 16. c3 {0.43/26 8} Qe7 {0.41/24 13 (Rc8)}
17. Qa4 {0.48/24 9 (Nc2)} Re8 {0.52/25 21} 18. Nd3 {0.56/25 4} Qc7 {0.45/23 5}
19. Nf4 {0.54/24 11} Qb6 {0.44/24 7} 20. Qa3 {0.53/24 5} a6 {0.49/23 7 (Qd8)}
21. Re1 {0.44/26 7} Ne7 {0.53/24 12} 22. g4 {0.80/24 9 (Bd2)} Bh7 {0.70/23 5
(Bd7)} 23. Ng2 {0.76/23 6} Qc6 {0.74/24 18 (Nc8)} 24. f3 {0.79/24 5} Nc8 {0.76/
23 3} 25. Bd2 {0.86/26 16 (Bf4)} Nb6 {0.81/24 4} 26. b3 {0.83/25 4} Na8 {0.82/
26 8 (Nbd7)} 27. Bf1 {0.79/24 5} Rc8 {0.80/25 13 (Nc7)} 28. Nf4 {0.96/23 6} Nc7
{0.95/23 2} 29. h4 {0.91/23 4} Re8 {0.96/23 3} 30. Rxe8+ {0.91/25 14} Nfxe8 {
0.86/24 2} 31. Kf2 {0.81/24 8} Ne6 {0.88/24 12} 32. Nxe6 {0.88/24 5} fxe6 {0.
80/23 2} 33. c4 {0.72/24 5} Qb6 {0.80/24 5} 34. Be3 {0.78/23 4} Qc7 {0.68/23 7
(Qd8)} 35. Qa4 {0.82/22 4} Qe7 {0.74/22 3} 36. g5 {0.74/22 6 (Kg2)} Qf7 {0.73/
23 12 (Bf5)} 37. Kg3 {0.64/21 8 (gxh6)} Bf5 {0.73/22 5} 38. Bf4 {0.50/21 4}
dxc4 {0.76/23 5 (Qh5)} 39. bxc4 {0.53/24 11} Qg6 {0.74/23 7} 40. Qb3 {0.58/23
7 (Bd2)} Nf6 {0.48/23 4 (hxg5)} 41. Kf2 {0.61/22 2 (Bd2)} Nh5 {0.49/26 9 (hxg5)
} 42. Bd2 {0.39/23 4} hxg5 {0.58/24 7} 43. hxg5 {0.44/23 2} Qf7 {0.55/23 2} 44.
Qe3 {0.50/23 4} Qd7 {0.56/24 6 (Qc7)} 45. a3 {0.43/24 5} g6 {0.34/23 3 (Kh7)}
46. Ba5 {0.37/21 2 (c5)} Kf7 {0.36/22 3} 47. Bb4 {0.35/21 2} Qc6 {0.36/21 1
(Qd8)} 48. Bd2 {0.36/22 2 (a4)} b6 {0.29/21 1 (Kg8)} 49. d5 {0.29/23 2} Qc5 {
0.36/25 4} 50. Bb4 {0.32/27 2} Qxe3+ {0.28/26 1} 51. Kxe3 {0.29/29 1} Ke7 {0.
31/29 3} 52. f4 {0.29/31 2 (Be1)} Ng3 {0.19/30 2} 53. Bg2 {0.29/31 1} exd5 {0.
19/32 2} 54. Kf3 {0.29/32 2 (Be1)} Ne4 {0.19/33 2} 55. cxd5 {0.29/33 1} b5 {0.
19/35 2} 56. Ke3 {0.29/34 1} Nc5 {0.19/35 1} 57. Bf3 {0.28/34 1} Kd7 {0.22/34
3 (Nd3)} 58. Bc3 {0.28/31 1 (Bd1)} Kc7 {0.21/34 6} 59. Ba5+ {0.28/36 2} Kc8 {
0.21/31 1 (Kd7)} 60. Bg2 {0.28/32 1 (Be1)} Kd7 {0.22/31 1} 61. Bc3 {0.28/37 2
(Bf1)} Kc7 {0.22/31 1 (Ke7)} 62. Bf3 {0.28/33 2} Kb6 {0.21/32 1 (Kd7)} 63. Bb4
{0.24/31 2 (Be2)} a5 {0.20/30 1} 64. Bc3 {0.21/30 2} Ka6 {0.20/32 2} 65. Be2 {
0.21/31 2 (Bd1)} Ne4 {0.20/31 2 (Kb6)} 66. Be1 {0.22/30 2} Nc5 {0.21/29 1} 67.
Bf1 {0.16/32 6 (Bf3)} Kb6 {0.20/29 2} 68. Be2 {0.15/32 2 (Bg2)} Ka6 {0.20/32 2}
69. Bf2 {0.14/31 3 (Bd1)} Kb6 {0.19/31 2} 70. Kd4 {0.04/29 2} Bc2 {0.18/28 1
(Nb3+)} 71. Be3 {0.04/28 4 (Be1)} Bf5 {0.17/31 3} 72. Bg1 {0.04/29 1 (Bf3)} Be4
{0.11/32 7} 73. Bd1 {0.03/30 2} Ka6 {0.07/27 1 (Bf5)} 74. Be3 {0.00/29 2} Bf5 {
0.06/28 1} 75. Bg1 {0.00/31 2 (Bf3)} Be4 {0.03/26 1} 76. Be3 {0.00/31 1 (Kc3)}
Bf5 {0.05/30 2} 77. Bf3 {0.00/33 1} Kb7 {0.05/31 2} 78. Kc3 {0.00/32 2 (Bg2)}
Na4+ {0.00/38 3} 79. Kd4 {0.00/36 1} Nc5 {0.00/28 1} 80. Be2 {0.00/36 2 (Bg1)}
Kb6 {0.00/29 1} 81. Kc3 {0.00/32 1 (Bf3)} Be4 {0.00/31 1} 82. Bg4 {0.00/32 1
(Kd4)} Bxd5 {0.00/28 1} 83. f5 {0.00/33 1 Draw accepted} 1/2-1/2

[mjlef]

Just a reminder that some newer AMD processors run the BMI2 PEXT instruction slower than normal. So pn these machine, just use the regular 64 bit version instead of the BMI2 version. I am not sure why the PEXT instruction is so slow on some of these chips. It would be nice if a microcode update told fix that, but I am guessing they are emulate PEXT in software.

[Werewolf]

Mark,

Am I right in saying the BMI2 build will benefit Haswell onwards?

I thought we once had a conversation together where you thought Skylake had removed the instructions to take advantage of this. Is that correct?

If so, is the situation the same with Kaby Lake and Coffee Lake etc?

Thanks, Carl

[mjlef]

Mark,

Am I right in saying the BMI2 build will benefit Haswell onwards?

I thought we once had a conversation together where you thought Skylake had removed the instructions to take advantage of this. Is that correct?

If so, is the situation the same with Kaby Lake and Coffee Lake etc?

Thanks, Carl

So these support BMI2:

"Intel Haswell processors and newer (like Skylake, Broadwell)"

I cannot predict what Intel might do in the future, but so far Intel has continued support in these lines.

I have a small program to test some CPU bit to see what is supported. I might put that on our website so people can run it on their machines to see what is supported. But basically, if the BMI2 compile does not work, use the other 64 bit version. For AMD, use the non-BMI2 version. At least until they speed up the BMI2 instructions.

[tpoppins]

So these support BMI2:

"Intel Haswell processors and newer (like Skylake, Broadwell)"

I cannot predict what Intel might do in the future, but so far Intel has continued support in these lines.

Although BMI2 instructions give only up to 5% speedup in some applications itt would make zero sense business-wise for Intel to remove them from their new processors unless it were to replace them with something considerably better (which is not likely to happen for a long while yet).

I have a small program to test some CPU bit to see what is supported. I might put that on our website so people can run it on their machines to see what is supported.

Or you could simply post a link to CPU-Z which is the most common tool for this purpose on Windoze and not have to provide support for yet another program. :)