Dev Builds » 20211207-1708

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 06:52:19 584987 3990 782 1174 2034 -34.24 ± 5.26 7 679 1008 301 0 -67.9 ± 10.71
ncm-dbt-02 06:53:18 586784 3996 833 1157 2006 -28.23 ± 5.23 3 639 1040 311 5 -57.2 ± 10.54
ncm-dbt-03 06:53:08 586653 4016 850 1224 1942 -32.45 ± 5.14 7 653 1055 293 0 -64.22 ± 10.45
ncm-dbt-04 06:52:55 571571 4002 863 1227 1912 -31.69 ± 5.29 10 659 1018 313 1 -62.3 ± 10.66
ncm-dbt-05 06:51:50 584543 3996 868 1202 1926 -29.11 ± 5.23 10 626 1053 306 3 -57.38 ± 10.46
20000 4196 5984 9820 -31.14 ± 2.34 37 3256 5174 1524 9 -61.79 ± 4.72

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
399779 ncm-dbt-04 572679 2 0 1 1 -189.7 ± 55.98 0 1 0 0 0 -1129.65 ± 376.02
399778 ncm-dbt-03 585000 16 4 4 8 0.0 ± 60.91 0 1 6 1 0 -0.0 ± 125.57
399777 ncm-dbt-01 585970 490 104 135 251 -22.01 ± 14.71 1 70 133 41 0 -42.76 ± 29.44
399776 ncm-dbt-05 585379 496 105 153 238 -33.73 ± 14.85 1 83 127 37 0 -66.65 ± 30.24
399775 ncm-dbt-02 585970 496 118 149 229 -21.74 ± 15.55 1 77 123 46 1 -43.66 ± 30.78
399774 ncm-dbt-04 571431 500 98 150 252 -36.26 ± 15.17 4 79 133 33 1 -68.99 ± 29.43
399773 ncm-dbt-03 586308 500 108 157 235 -34.16 ± 15.16 2 84 125 39 0 -66.1 ± 30.52
399772 ncm-dbt-01 581985 500 95 147 258 -36.26 ± 14.91 1 87 125 37 0 -71.89 ± 30.51
399771 ncm-dbt-05 585632 500 125 160 215 -24.36 ± 13.97 1 69 144 36 0 -47.55 ± 28.01
399770 ncm-dbt-02 588941 500 103 138 259 -24.36 ± 15.26 0 82 121 47 0 -48.96 ± 31.04
399769 ncm-dbt-03 586774 500 116 147 237 -21.57 ± 14.68 0 75 131 44 0 -43.3 ± 29.76
399768 ncm-dbt-04 570989 500 110 148 242 -26.45 ± 15.52 1 83 119 47 0 -51.8 ± 31.29
399767 ncm-dbt-05 587962 500 99 143 258 -30.65 ± 14.29 1 77 137 35 0 -60.36 ± 28.92
399766 ncm-dbt-01 584874 500 97 159 244 -43.3 ± 15.45 1 98 113 38 0 -86.52 ± 32.11
399765 ncm-dbt-02 587325 500 111 164 225 -36.97 ± 15.46 0 94 116 39 1 -76.25 ± 31.7
399764 ncm-dbt-03 587792 500 100 159 241 -41.19 ± 13.76 0 86 137 27 0 -83.57 ± 28.79
399763 ncm-dbt-04 572316 500 98 154 248 -39.08 ± 14.57 1 87 129 33 0 -77.71 ± 29.96
399762 ncm-dbt-05 579001 500 96 159 245 -44.01 ± 14.47 1 90 131 27 1 -89.48 ± 29.62
399761 ncm-dbt-01 585000 500 97 138 265 -28.55 ± 14.01 0 76 139 35 0 -57.5 ± 28.65
399760 ncm-dbt-02 589283 500 95 147 258 -36.26 ± 13.15 0 77 148 25 0 -73.34 ± 27.25
399759 ncm-dbt-03 586308 500 104 164 232 -41.89 ± 14.74 2 88 128 32 0 -82.1 ± 30.08
399758 ncm-dbt-04 571471 500 104 157 239 -36.97 ± 14.96 1 88 124 37 0 -73.34 ± 30.64
399757 ncm-dbt-05 583070 500 111 146 243 -24.36 ± 16.21 2 82 117 47 2 -48.96 ± 31.54
399756 ncm-dbt-01 584706 500 97 137 266 -27.85 ± 15.49 1 84 119 46 0 -54.65 ± 31.29
399755 ncm-dbt-02 585084 500 91 141 268 -34.86 ± 14.17 0 83 134 33 0 -70.44 ± 29.29
399754 ncm-dbt-03 585126 500 96 145 259 -34.16 ± 14.52 1 82 132 35 0 -67.55 ± 29.58
399753 ncm-dbt-04 570629 500 115 163 222 -33.46 ± 15.73 0 94 110 46 0 -67.55 ± 32.45
399752 ncm-dbt-05 585632 500 114 144 242 -20.87 ± 14.63 3 65 141 41 0 -37.67 ± 28.45
399751 ncm-dbt-01 584622 500 84 149 267 -45.42 ± 14.55 2 90 129 29 0 -89.48 ± 29.91
399750 ncm-dbt-02 588047 500 105 138 257 -22.97 ± 15.16 0 79 126 44 1 -47.55 ± 30.4
399749 ncm-dbt-03 585928 500 95 148 257 -36.97 ± 14.7 1 86 128 35 0 -73.34 ± 30.1
399748 ncm-dbt-04 569031 500 101 155 244 -37.67 ± 15.5 2 90 118 40 0 -73.34 ± 31.44
399747 ncm-dbt-05 583531 500 110 144 246 -23.66 ± 14.96 1 76 129 44 0 -46.13 ± 30.02
399746 ncm-dbt-01 585885 500 95 157 248 -43.3 ± 14.82 0 96 120 34 0 -88.0 ± 31.17
399745 ncm-dbt-03 588174 500 120 159 221 -27.15 ± 14.31 1 74 138 37 0 -53.22 ± 28.81
399744 ncm-dbt-02 584874 500 102 135 263 -22.97 ± 14.13 0 71 142 36 1 -47.55 ± 28.28
399743 ncm-dbt-04 572316 500 113 163 224 -34.86 ± 13.63 0 79 142 29 0 -70.44 ± 28.16
399742 ncm-dbt-05 586139 500 108 153 239 -31.35 ± 14.73 0 84 127 39 0 -63.23 ± 30.26
399741 ncm-dbt-01 586858 500 113 152 235 -27.15 ± 14.83 1 78 130 41 0 -53.22 ± 29.88
399740 ncm-dbt-03 588472 500 107 141 252 -23.66 ± 14.71 0 77 130 43 0 -47.55 ± 29.89
399739 ncm-dbt-02 584748 500 108 145 247 -25.76 ± 15.23 2 76 130 41 1 -50.38 ± 29.88
399738 ncm-dbt-04 573284 500 124 136 240 -8.34 ± 14.26 1 58 143 48 0 -15.3 ± 28.23

Commit

Commit ID 4766dfc3956f78d853c5e0c4636d6f90fd93df9a
Author Tomasz Sobczyk
Date 2021-12-07 17:08:54 UTC
Optimize FT activation and affine transform for NEON. This patch optimizes the NEON implementation in two ways. The activation layer after the feature transformer is rewritten to make it easier for the compiler to see through dependencies and unroll. This in itself is a minimal, but a positive improvement. Other architectures could benefit from this too in the future. This is not an algorithmic change. The affine transform for large matrices (first layer after FT) on NEON now utilizes the same optimized code path as >=SSSE3, which makes the memory accesses more sequential and makes better use of the available registers, which allows for code that has longer dependency chains. Benchmarks from Redshift#161, profile-build with apple clang george@Georges-MacBook-Air nets % ./stockfish-b82d93 bench 2>&1 | tail -4 (current master) =========================== Total time (ms) : 2167 Nodes searched : 4667742 Nodes/second : 2154011 george@Georges-MacBook-Air nets % ./stockfish-7377b8 bench 2>&1 | tail -4 (this patch) =========================== Total time (ms) : 1842 Nodes searched : 4667742 Nodes/second : 2534061 This is a solid 18% improvement overall, larger in a bench with NNUE-only, not mixed. Improvement is also observed on armv7-neon (Raspberry Pi, and older phones), around 5% speedup. No changes for architectures other than NEON. closes https://github.com/official-stockfish/Stockfish/pull/3837 No functional changes.
Copyright 2011–2025 Next Chess Move LLC