Dev Builds » 20211203-0751

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 14. 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 09:56:33 1145071 3344 1129 542 1673 +61.63 +/- 5.66 2 149 798 706 17 +123.85 +/- 12.02
ncm-dbt-02 09:53:02 1264140 3324 1145 523 1656 +65.79 +/- 5.8 0 154 756 728 24 +130.87 +/- 12.38
ncm-dbt-03 09:56:48 1272727 3338 1135 564 1639 +60.02 +/- 5.86 1 180 753 717 18 +119.87 +/- 12.41
ncm-dbt-04 09:57:15 1255005 3346 1155 535 1656 +65.13 +/- 5.78 1 161 745 749 17 +131.35 +/- 12.48
ncm-dbt-05 09:53:09 1264379 3312 1112 515 1685 +63.32 +/- 5.82 6 150 757 727 16 +128.74 +/- 12.37
ncm-dbt-06 09:57:09 1258364 3336 1109 541 1686 +59.74 +/- 5.75 3 169 762 725 9 +121.82 +/- 12.34
20000 6785 3220 9995 +62.6 +/- 2.36 13 963 4571 4352 101 +126.06 +/- 5.03

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
206574 ncm-dbt-05 1223434 144 46 23 75 +55.97 +/- 24.63 0 5 39 28 0 +115.01 +/- 53.91
206573 ncm-dbt-02 1237271 162 57 26 79 +67.31 +/- 27.36 0 9 33 38 1 +135.1 +/- 60.01
206572 ncm-dbt-03 1251930 170 66 34 70 +66.18 +/- 24.95 0 7 40 37 1 +132.83 +/- 54.25
206571 ncm-dbt-06 1215126 172 60 22 90 +78.04 +/- 24.03 0 6 36 44 0 +164.87 +/- 57.53
206570 ncm-dbt-01 1086892 172 58 27 87 +63.31 +/- 25.88 0 9 38 38 1 +126.51 +/- 55.88
206569 ncm-dbt-04 1220018 180 57 31 92 +50.53 +/- 26.4 0 13 39 37 1 +99.11 +/- 54.98
206568 ncm-dbt-05 1231550 500 171 73 256 +68.99 +/- 15.11 0 23 110 113 4 +137.37 +/- 32.59
206567 ncm-dbt-02 1236642 500 161 74 265 +61.07 +/- 15.18 0 28 109 111 2 +123.02 +/- 32.74
206566 ncm-dbt-03 1236587 500 174 81 245 +65.38 +/- 15.1 0 24 113 109 4 +129.35 +/- 32.14
206565 ncm-dbt-06 1222609 500 159 87 254 +50.38 +/- 14.28 0 27 124 99 0 +102.97 +/- 30.57
206564 ncm-dbt-04 1246628 500 178 83 239 +66.82 +/- 13.48 0 14 129 105 2 +135.76 +/- 29.49
206563 ncm-dbt-01 1085602 500 170 71 259 +69.71 +/- 15.66 1 23 107 114 5 +138.99 +/- 33.06
206562 ncm-dbt-05 1233942 500 167 84 249 +58.21 +/- 15.32 2 23 118 104 3 +118.33 +/- 31.4
206561 ncm-dbt-02 1228264 500 166 84 250 +57.5 +/- 14.63 0 25 120 103 2 +115.23 +/- 31.11
206560 ncm-dbt-03 1241364 500 173 80 247 +65.38 +/- 15.62 0 29 102 116 3 +130.94 +/- 33.79
206559 ncm-dbt-06 1213118 500 158 77 265 +56.78 +/- 14.32 0 24 122 103 1 +115.23 +/- 30.81
206558 ncm-dbt-01 1100510 500 175 86 239 +62.51 +/- 14.17 0 21 120 108 1 +127.76 +/- 31.05
206557 ncm-dbt-04 1203975 500 170 78 252 +64.66 +/- 16.82 0 37 89 119 5 +126.18 +/- 35.6
206556 ncm-dbt-05 1213937 500 153 85 262 +47.55 +/- 15.68 1 33 116 97 3 +93.95 +/- 31.72
206555 ncm-dbt-02 1229531 500 170 87 243 +58.21 +/- 14.53 0 22 127 97 4 +113.68 +/- 30.06
206554 ncm-dbt-03 1230742 500 163 78 259 +59.64 +/- 14.18 0 22 122 105 1 +121.46 +/- 30.78
206553 ncm-dbt-04 1219747 500 172 90 238 +57.5 +/- 14.76 0 26 118 104 2 +115.23 +/- 31.41
206552 ncm-dbt-06 1232237 500 184 88 228 +67.54 +/- 14.92 0 24 108 116 2 +137.37 +/- 32.9
206551 ncm-dbt-01 1099489 500 164 81 255 +58.21 +/- 13.97 0 19 132 96 3 +115.23 +/- 29.26
206550 ncm-dbt-05 1237605 500 186 78 236 +76.25 +/- 14.39 0 16 114 116 4 +153.86 +/- 31.85
206549 ncm-dbt-02 1239608 500 169 76 255 +65.38 +/- 14.42 0 18 126 101 5 +127.76 +/- 30.09
206548 ncm-dbt-03 1243096 500 166 93 241 +51.09 +/- 14.98 0 31 116 102 1 +102.97 +/- 31.72
206547 ncm-dbt-04 1199031 500 180 76 244 +73.33 +/- 14.99 0 22 105 120 3 +148.85 +/- 33.38
206546 ncm-dbt-01 1100035 500 175 79 246 +67.55 +/- 14.37 1 16 122 108 3 +137.37 +/- 30.65
206545 ncm-dbt-06 1229856 500 166 82 252 +58.93 +/- 14.69 0 27 112 111 0 +121.46 +/- 32.29
206544 ncm-dbt-04 1230342 500 174 83 243 +63.95 +/- 14.64 1 21 115 112 1 +132.54 +/- 31.82
206543 ncm-dbt-05 1238846 500 174 68 258 +74.79 +/- 14.63 1 17 109 121 2 +155.54 +/- 32.7
206542 ncm-dbt-02 1219995 500 193 78 229 +81.36 +/- 15.24 0 19 103 122 6 +162.35 +/- 33.71
206541 ncm-dbt-03 1238281 500 171 85 244 +60.36 +/- 15.02 0 25 118 103 4 +118.33 +/- 31.4
206540 ncm-dbt-01 1109231 500 166 81 253 +59.64 +/- 14.59 0 23 122 102 3 +118.33 +/- 30.8
206539 ncm-dbt-06 1234682 500 163 81 256 +57.5 +/- 15.03 2 22 120 104 2 +118.33 +/- 31.1
206538 ncm-dbt-03 1246236 500 168 85 247 +58.21 +/- 16.19 1 32 104 109 4 +115.23 +/- 33.43
206537 ncm-dbt-02 1228406 500 168 74 258 +66.1 +/- 15.13 0 25 109 113 3 +132.54 +/- 32.75
206536 ncm-dbt-05 1230681 500 164 85 251 +55.36 +/- 15.19 2 26 113 109 0 +116.78 +/- 32.15
206535 ncm-dbt-06 1222449 500 166 78 256 +61.79 +/- 15.85 1 30 101 116 2 +126.18 +/- 33.92
206534 ncm-dbt-01 1101434 500 162 79 259 +58.21 +/- 14.93 0 28 112 109 1 +118.33 +/- 32.29
206533 ncm-dbt-04 1233715 500 170 66 264 +73.33 +/- 14.72 0 20 109 118 3 +148.85 +/- 32.73
177697 ncm-dbt-02 1493407 162 61 24 77 +80.77 +/- 27.44 0 8 29 43 1 +165.99 +/- 64.0
177696 ncm-dbt-04 1486586 166 54 28 84 +54.87 +/- 24.51 0 8 41 34 0 +112.62 +/- 53.47
177695 ncm-dbt-05 1505041 168 51 19 98 +66.99 +/- 24.5 0 7 38 39 0 +139.38 +/- 55.82
177694 ncm-dbt-01 1477381 172 59 38 75 +42.63 +/- 24.11 0 10 45 31 0 +86.59 +/- 50.86
177693 ncm-dbt-06 1496842 164 53 26 85 +57.72 +/- 27.09 0 9 39 32 2 +109.4 +/- 55.04
177692 ncm-dbt-03 1493581 168 54 28 86 +54.2 +/- 25.64 0 10 38 36 0 +111.19 +/- 55.85

Commit

Commit ID c9977aa0a89c83bf21651bffd3b6f10c344ccc46
Author Gian-Carlo Pascutto
Date 2021-12-03 07:51:06 UTC
Add AVX-VNNI support for Alder Lake and later. In their infinite wisdom, Intel axed AVX512 from Alder Lake chips (well, not entirely, but we kind of want to use the Gracemont cores for chess!) but still added VNNI support. Confusingly enough, this is not the same as VNNI256 support. This adds a specific AVX-VNNI target that will use this AVX-VNNI mode, by prefixing the VNNI instructions with the appropriate VEX prefix, and avoiding AVX512 usage. This is about 1% faster on P cores: Result of 20 runs ================== base (./clang-bmi2 ) = 3306337 +/- 7519 test (./clang-vnni ) = 3344226 +/- 7388 diff = +37889 +/- 4153 speedup = +0.0115 P(speedup > 0) = 1.0000 But a nice 3% faster on E cores: Result of 20 runs ================== base (./clang-bmi2 ) = 1938054 +/- 28257 test (./clang-vnni ) = 1994606 +/- 31756 diff = +56552 +/- 3735 speedup = +0.0292 P(speedup > 0) = 1.0000 This was measured on Clang 13. GCC 11.2 appears to generate worse code for Alder Lake, though the speedup on the E cores is similar. It is possible to run the engine specifically on the P or E using binding, for example in linux it is possible to use (for an 8 P + 8 E setup like i9-12900K): taskset -c 0-15 ./stockfish taskset -c 16-23 ./stockfish where the first call binds to the P-cores and the second to the E-cores. closes https://github.com/official-stockfish/Stockfish/pull/3824 No functional change
Copyright 2011–2024 Next Chess Move LLC