Dev Builds » 20240523-1937

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 15. This yields an approximate Elo difference and establishes confidence in the strength of the dev builds.

Summary

Host Duration Avg Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo
ncm-dbt-01 06:51:06 583249 4000 1498 663 1839 +73.61 ± 5.0 1 136 903 947 13 +151.97 ± 11.29
ncm-dbt-02 06:53:11 586885 4010 1463 642 1905 +72.15 ± 5.02 4 141 896 958 6 +150.71 ± 11.35
ncm-dbt-03 06:52:55 587511 4000 1487 635 1878 +75.15 ± 4.91 1 120 917 950 12 +155.76 ± 11.18
ncm-dbt-04 06:53:26 571437 4004 1481 683 1840 +70.18 ± 5.04 2 147 916 925 12 +144.55 ± 11.22
ncm-dbt-05 06:54:24 583322 3986 1440 628 1918 +71.78 ± 5.12 2 156 873 952 10 +148.6 ± 11.52
20000 7369 3251 9380 +72.57 ± 2.25 10 700 4505 4732 53 +150.3 ± 5.06

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
385954 ncm-dbt-04 569869 4 2 1 1 +88.51 ± 137.63 0 0 1 1 0 +190.5 ± 634.21
385953 ncm-dbt-02 585337 10 5 4 1 +34.8 ± 62.53 0 0 4 1 0 +70.35 ± 134.03
385952 ncm-dbt-05 578547 486 180 64 242 +84.56 ± 14.15 0 15 97 131 0 +180.52 ± 34.74
385951 ncm-dbt-01 582235 500 186 78 236 +76.25 ± 13.96 0 15 114 119 2 +157.24 ± 31.83
385950 ncm-dbt-03 586604 500 189 72 239 +82.83 ± 14.14 0 15 105 128 2 +172.78 ± 33.32
385949 ncm-dbt-02 584537 500 185 84 231 +71.16 ± 13.49 0 15 119 116 0 +148.85 ± 31.04
385948 ncm-dbt-04 569869 500 189 93 218 +67.54 ± 15.05 0 22 115 108 5 +132.54 ± 31.82
385947 ncm-dbt-05 583531 500 175 92 233 +58.21 ± 14.11 0 22 124 103 1 +118.33 ± 30.49
385946 ncm-dbt-01 585042 500 181 74 245 +75.52 ± 14.23 0 18 108 123 1 +157.24 ± 32.87
385945 ncm-dbt-02 588686 500 173 69 258 +73.33 ± 13.86 0 16 115 118 1 +152.18 ± 31.7
385944 ncm-dbt-03 586774 500 181 80 239 +71.16 ± 13.93 0 18 113 119 0 +148.85 ± 32.06
385943 ncm-dbt-04 571712 500 184 97 219 +61.07 ± 14.24 0 22 120 107 1 +124.6 ± 31.07
385942 ncm-dbt-01 586943 500 184 90 226 +66.1 ± 13.15 0 14 128 108 0 +137.37 ± 29.63
385941 ncm-dbt-05 588089 500 191 80 229 +78.43 ± 14.74 0 20 101 127 2 +162.35 ± 34.05
385940 ncm-dbt-04 572517 500 188 86 226 +71.88 ± 13.81 0 16 117 116 1 +148.85 ± 31.38
385939 ncm-dbt-02 587409 500 181 79 240 +71.89 ± 14.1 2 13 116 119 0 +153.86 ± 31.51
385938 ncm-dbt-03 588643 500 197 85 218 +79.17 ± 13.29 0 12 114 124 0 +167.53 ± 31.71
385937 ncm-dbt-01 581985 500 198 83 219 +81.36 ± 13.81 0 13 111 124 2 +169.27 ± 32.25
385936 ncm-dbt-05 586054 500 186 88 226 +68.99 ± 14.7 2 18 110 120 0 +147.19 ± 32.57
385935 ncm-dbt-04 572679 500 192 94 214 +68.99 ± 14.42 2 14 120 112 2 +143.89 ± 30.92
385934 ncm-dbt-02 587962 500 187 87 226 +70.43 ± 13.91 0 16 120 112 2 +143.89 ± 30.92
385933 ncm-dbt-03 587494 500 178 71 251 +75.52 ± 14.79 1 17 109 120 3 +155.54 ± 32.7
385932 ncm-dbt-05 586350 500 184 81 235 +72.61 ± 14.41 0 20 108 121 1 +150.51 ± 32.89
385931 ncm-dbt-02 586562 500 183 86 231 +68.27 ± 14.81 1 22 106 121 0 +143.89 ± 33.22
385930 ncm-dbt-01 584537 500 201 85 214 +82.1 ± 14.7 0 17 104 125 4 +167.53 ± 33.52
385929 ncm-dbt-04 571471 500 187 72 241 +81.36 ± 13.66 0 13 110 126 1 +171.02 ± 32.42
385928 ncm-dbt-03 587409 500 193 85 222 +76.25 ± 14.1 0 15 115 117 3 +155.54 ± 31.67
385927 ncm-dbt-01 582903 500 180 72 248 +76.25 ± 13.96 0 16 111 122 1 +158.93 ± 32.34
385926 ncm-dbt-02 588047 500 190 90 220 +70.44 ± 14.2 1 17 113 119 0 +148.85 ± 32.06
385925 ncm-dbt-04 570789 500 184 84 232 +70.43 ± 14.34 0 20 111 118 1 +145.54 ± 32.41
385924 ncm-dbt-05 584537 500 172 64 264 +76.25 ± 14.25 0 19 104 127 0 +160.64 ± 33.54
385923 ncm-dbt-03 586901 500 187 84 229 +72.61 ± 13.84 0 14 122 111 3 +147.19 ± 30.53
385922 ncm-dbt-01 582527 500 187 94 219 +65.38 ± 14.42 0 22 114 113 1 +134.15 ± 31.97
385921 ncm-dbt-02 588132 500 178 72 250 +74.79 ± 14.34 0 19 107 123 1 +155.54 ± 33.04
385920 ncm-dbt-04 571511 500 184 77 239 +75.52 ± 13.79 0 15 114 120 1 +157.24 ± 31.83
385919 ncm-dbt-03 588728 500 182 84 234 +68.99 ± 13.11 0 13 126 111 0 +143.89 ± 29.88
385918 ncm-dbt-05 582318 500 178 82 240 +67.54 ± 14.51 0 22 111 116 1 +138.99 ± 32.43
385917 ncm-dbt-01 579826 500 181 87 232 +66.1 ± 14.86 1 21 113 113 2 +135.76 ± 32.12
385916 ncm-dbt-02 585295 500 181 71 248 +77.7 ± 15.13 0 23 96 129 2 +160.64 ± 34.88
385915 ncm-dbt-03 587537 500 180 74 246 +74.79 ± 13.91 0 16 113 120 1 +155.54 ± 32.01
385914 ncm-dbt-04 572517 500 171 79 250 +64.66 ± 14.67 0 25 108 117 0 +134.15 ± 32.9
385913 ncm-dbt-05 577151 500 174 77 249 +68.27 ± 14.81 0 20 118 107 5 +134.15 ± 31.33

Commit

Commit ID c6a1e7fd4232ec151206fab16cb7daa23bfd7137
Author cj5716
Date 2024-05-23 19:37:46 UTC
Optimise pairwise multiplication This speedup was first inspired by a comment by @AndyGrant on my recent PR "If mullo_epi16 would preserve the signedness, then this could be used to remove 50% of the max operations during the halfkp-pairwise mat-mul relu deal." That got me thinking, because although mullo_epi16 did not preserve the signedness, mulhi_epi16 did, and so we could shift left and then use mulhi_epi16, instead of shifting right after the mullo. However, due to some issues with shifting into the sign bit, the FT weights and biases had to be multiplied by 2 for the optimisation to work. Speedup on "Arch=x86-64-bmi2 COMP=clang", courtesy of @Torom Result of 50 runs base (...es/stockfish) = 962946 +/- 1202 test (...ise-max-less) = 979696 +/- 1084 diff = +16750 +/- 1794 speedup = +0.0174 P(speedup > 0) = 1.0000 CPU: 4 x Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz Hyperthreading: on Also a speedup on "COMP=gcc", courtesy of Torom once again Result of 50 runs base (...tockfish_gcc) = 966033 +/- 1574 test (...max-less_gcc) = 983319 +/- 1513 diff = +17286 +/- 2515 speedup = +0.0179 P(speedup > 0) = 1.0000 CPU: 4 x Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz Hyperthreading: on Passed STC: LLR: 2.96 (-2.94,2.94) <0.00,2.00> Total: 67712 W: 17715 L: 17358 D: 32639 Ptnml(0-2): 225, 7472, 18140, 7759, 260 https://tests.stockfishchess.org/tests/view/664c1d75830eb9f886616906 closes https://github.com/official-stockfish/Stockfish/pull/5282 No functional change
Copyright 2011–2024 Next Chess Move LLC