Dev Builds » 20230218-1223

You are viewing an old NCM Stockfish dev build test. You may find the most recent dev build tests using Stockfish 15 as the baseline here.

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 05:00:52 1117582 1670 691 179 800 +110.06 ± 7.93 1 33 274 507 20 +235.67 ± 20.62
ncm-dbt-02 04:58:44 1235829 1652 674 159 819 +112.04 ± 7.59 0 25 277 508 16 +243.07 ± 20.47
ncm-dbt-03 05:00:03 1240707 1684 680 178 826 +106.81 ± 7.46 0 25 305 497 15 +229.32 ± 19.46
ncm-dbt-04 05:02:00 1235170 1660 681 164 815 +111.92 ± 7.83 1 26 280 501 22 +239.52 ± 20.37
ncm-dbt-05 04:58:43 1221766 1662 687 168 807 +112.24 ± 7.64 1 24 278 511 17 +243.71 ± 20.44
ncm-dbt-06 05:01:00 1228538 1672 671 181 820 +104.89 ± 7.91 0 40 282 498 16 +223.43 ± 20.32
ncm-et-3 06:25:33 1304438 1668 685 170 813 +110.89 ± 7.83 0 30 281 501 22 +236.05 ± 20.35
ncm-et-4 06:25:07 1299957 1670 687 152 831 +115.37 ± 7.87 1 32 252 531 19 +251.42 ± 21.5
ncm-et-9 06:25:48 1298583 1652 696 175 781 +113.44 ± 8.25 2 34 257 507 26 +241.75 ± 21.29
ncm-et-10 06:26:17 1289676 1672 654 179 839 +101.49 ± 7.77 0 36 304 481 15 +214.97 ± 19.55
ncm-et-13 06:26:06 1305921 1670 668 186 816 +103.21 ± 7.81 2 32 297 490 14 +221.33 ± 19.78
ncm-et-15 06:25:23 1305980 1668 675 172 821 +108.13 ± 7.46 0 26 291 505 12 +234.77 ± 19.96
20000 8149 2063 9788 +109.18 ± 2.25 8 363 3378 6037 214 +234.38 ± 5.86

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
201401 ncm-dbt-02 1228645 152 58 20 74 +88.73 ± 29.15 0 7 26 41 2 +178.86 ± 67.72
201400 ncm-dbt-04 1241665 160 69 19 72 +112.32 ± 27.88 0 4 26 46 4 +227.56 ± 68.48
201399 ncm-dbt-05 1244027 162 67 18 77 +108.48 ± 24.25 0 1 33 44 3 +223.89 ± 59.15
201398 ncm-dbt-06 1248173 172 70 16 86 +112.88 ± 23.33 0 3 27 55 1 +249.8 ± 67.28
201397 ncm-dbt-01 1109430 170 73 17 80 +118.88 ± 25.92 0 3 27 51 4 +247.28 ± 67.26
201396 ncm-dbt-03 1236652 184 76 21 87 +107.12 ± 22.34 0 3 32 56 1 +233.83 ± 61.35
201395 ncm-dbt-02 1224489 500 203 41 256 +116.77 ± 13.19 0 4 85 156 5 +256.44 ± 36.95
201394 ncm-dbt-04 1248256 500 211 49 240 +116.77 ± 14.53 0 8 81 152 9 +247.41 ± 38.13
201393 ncm-dbt-01 1123242 500 204 61 235 +102.22 ± 14.63 0 12 89 143 6 +213.85 ± 36.34
201392 ncm-dbt-06 1231122 500 200 60 240 +99.95 ± 15.05 0 14 89 140 7 +206.01 ± 36.33
201391 ncm-dbt-05 1208010 500 194 50 256 +102.97 ± 12.87 0 5 98 145 2 +223.94 ± 34.18
201390 ncm-dbt-03 1234295 500 200 54 246 +104.49 ± 13.71 0 8 92 146 4 +223.94 ± 35.61
201389 ncm-dbt-02 1256222 500 213 52 235 +116.0 ± 14.05 0 8 79 157 6 +251.89 ± 38.63
201388 ncm-dbt-01 1124193 500 202 44 254 +113.68 ± 13.38 0 7 81 159 3 +251.89 ± 38.11
201387 ncm-dbt-04 1215536 500 203 47 250 +112.14 ± 13.56 0 7 84 155 4 +245.2 ± 37.37
201386 ncm-dbt-05 1220764 500 212 55 233 +112.91 ± 14.22 1 7 81 156 5 +247.41 ± 38.13
201385 ncm-dbt-03 1250735 500 201 55 244 +104.49 ± 13.05 0 4 100 142 4 +223.94 ± 33.69
201384 ncm-dbt-06 1213333 500 206 48 246 +113.68 ± 14.22 0 10 77 158 5 +247.41 ± 39.15
201383 ncm-dbt-03 1241149 500 203 48 249 +111.37 ± 14.38 0 10 81 153 6 +238.66 ± 38.15
201382 ncm-dbt-02 1233960 500 200 46 254 +110.6 ± 13.22 0 6 87 154 3 +243.0 ± 36.62
201381 ncm-dbt-05 1214265 500 214 45 241 +122.24 ± 14.66 0 11 66 166 7 +268.17 ± 42.21
201380 ncm-dbt-06 1221527 500 195 57 248 +98.44 ± 14.29 0 13 89 145 3 +209.91 ± 36.34
201379 ncm-dbt-01 1113463 500 212 57 231 +111.37 ± 15.15 1 11 77 154 7 +238.66 ± 39.1
201378 ncm-dbt-04 1235223 500 198 49 253 +106.78 ± 14.2 1 7 89 148 5 +230.16 ± 36.26
166628 ncm-et-9 1297277 152 61 17 74 +103.53 ± 24.27 0 3 26 47 0 +229.61 ± 68.49
166627 ncm-et-15 1310124 168 66 18 84 +102.1 ± 21.92 0 1 35 47 1 +219.63 ± 57.22
166626 ncm-et-4 1298927 170 71 12 87 +125.8 ± 25.03 0 4 20 59 2 +282.05 ± 78.34
166625 ncm-et-10 1291540 172 69 19 84 +103.99 ± 22.48 0 3 30 53 0 +230.89 ± 63.47
166624 ncm-et-13 1323911 170 61 13 96 +100.83 ± 24.24 0 4 30 50 1 +216.32 ± 63.49
166623 ncm-et-3 1308659 168 74 22 72 +111.18 ± 23.69 0 1 33 47 3 +231.91 ± 59.33
166622 ncm-et-9 1293526 500 208 49 243 +114.45 ± 15.6 1 10 79 149 11 +238.66 ± 38.63
166621 ncm-et-15 1305247 500 202 50 248 +109.07 ± 13.05 0 5 91 151 3 +238.66 ± 35.65
166620 ncm-et-4 1295469 500 210 42 248 +121.45 ± 14.5 0 9 72 161 8 +263.42 ± 40.52
166619 ncm-et-13 1287530 500 205 63 232 +101.46 ± 14.63 1 10 90 144 5 +215.85 ± 36.11
166618 ncm-et-10 1296573 500 195 56 249 +99.2 ± 15.47 0 15 90 136 9 +200.24 ± 36.12
166617 ncm-et-3 1309653 500 200 52 248 +106.01 ± 13.88 0 10 85 152 3 +230.16 ± 37.2
166616 ncm-et-9 1303059 500 219 51 230 +121.46 ± 14.98 1 8 72 160 9 +263.42 ± 40.52
166615 ncm-et-15 1302502 500 196 56 248 +99.95 ± 14.0 0 11 91 145 3 +213.85 ± 35.9
166614 ncm-et-13 1315647 500 209 55 236 +110.6 ± 14.22 0 10 81 154 5 +238.66 ± 38.15
166613 ncm-et-4 1300829 500 198 44 258 +110.6 ± 14.69 1 11 75 159 4 +243.0 ± 39.61
166612 ncm-et-3 1296067 500 202 54 244 +106.01 ± 14.82 0 11 88 143 8 +219.87 ± 36.54
166611 ncm-et-10 1289114 500 196 56 248 +99.95 ± 14.0 0 10 94 142 4 +211.87 ± 35.26
166610 ncm-et-15 1306047 500 211 48 241 +117.55 ± 14.04 0 9 74 162 5 +258.75 ± 39.96
166609 ncm-et-4 1304605 500 208 54 238 +110.6 ± 13.89 0 8 85 152 5 +238.66 ± 37.17
166608 ncm-et-13 1296597 500 193 55 252 +98.44 ± 13.98 1 8 96 142 3 +211.87 ± 34.83
166607 ncm-et-3 1303373 500 209 42 249 +120.67 ± 14.35 0 8 75 159 8 +261.07 ± 39.69
166606 ncm-et-9 1300473 500 208 58 234 +107.54 ± 14.83 0 13 80 151 6 +228.08 ± 38.34
166605 ncm-et-10 1281479 500 194 48 258 +104.49 ± 13.39 0 8 90 150 2 +228.08 ± 36.04

Commit

Commit ID 2c36d1e7e7374b8babb3cc503c0bc07ceb83dbf8
Author MinetaS
Date 2023-02-18 12:23:18 UTC
Fix overflow in add_dpbusd_epi32x2 This patch fixes 16bit overflow in *_add_dpbusd_epi32x2 functions, that can be triggered in rare cases depending on the NNUE weights. While the code leads to some slowdown on affected architectures (most notably avx2), the fix is simpler than some of the other options discussed in https://github.com/official-stockfish/Stockfish/pull/4394 Code suggested by Sopel97. Result of "bench 4096 1 30 default depth nnue": | Architecture | master | patch (gcc) | patch (clang) | |---------------------|-----------|-------------|---------------| | x86-64-vnni512 | 762122798 | 762122798 | 762122798 | | x86-64-avx512 | 769723503 | 762122798 | 762122798 | | x86-64-bmi2 | 769723503 | 762122798 | 762122798 | | x86-64-ssse3 | 769723503 | 762122798 | 762122798 | | x86-64 | 762122798 | 762122798 | 762122798 | Following architectures will experience ~4% slowdown due to an additional instruction in the middle of hot path: * x86-64-avx512 * x86-64-bmi2 * x86-64-avx2 * x86-64-sse41-popcnt (x86-64-modern) * x86-64-ssse3 * x86-32-sse41-popcnt This patch clearly loses Elo against master with both STC and LTC. Failed non-regression STC (256bit fix only): LLR: -2.95 (-2.94,2.94) <-1.75,0.25> Total: 33528 W: 8769 L: 9049 D: 15710 Ptnml(0-2): 96, 3616, 9600, 3376, 76 https://tests.stockfishchess.org/tests/view/63e6a5b44299542b1e26a485 60+0.6 @ 30000 games: Elo: -1.67 +-1.7 (95%) LOS: 2.8% Total: 30000 W: 7848 L: 7992 D: 14160 Ptnml(0-2): 12, 2847, 9436, 2683, 22 nElo: -3.84 +-3.9 (95%) PairsRatio: 0.95 https://tests.stockfishchess.org/tests/view/63e7ac716d0e1db55f35a660 However, a test against nn-a3dc078bafc7.nnue, which is the latest "safe" network not causing the bug, passed with regular bounds. Passed STC: LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 160456 W: 42658 L: 42175 D: 75623 Ptnml(0-2): 487, 17638, 43469, 18173, 461 https://tests.stockfishchess.org/tests/view/63e89836d62a5d02b0fa82c8 closes https://github.com/official-stockfish/Stockfish/pull/4391 closes https://github.com/official-stockfish/Stockfish/pull/4394 No functional change
Copyright 2011–2024 Next Chess Move LLC