Dev Builds » 20150910-1815

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 7. This yields an approximate Elo difference and establishes confidence in the strength of the dev builds.

Summary

Host Duration Avg Base NPS Games Wins Losses Draws Elo
ncm-et-3 05:55:47 1987477 2435 232 429 1774 -28.17 +/- 7.15
ncm-et-4 05:55:14 1975965 2430 205 425 1800 -31.54 +/- 6.98
ncm-et-5 01:49:15 2023773 749 73 134 542 -28.36 +/- 13.02
ncm-et-6 01:48:52 2028197 754 72 151 531 -36.54 +/- 13.39
ncm-et-7 01:49:03 2023534 766 70 128 568 -26.36 +/- 12.45
ncm-et-8 01:48:54 2016447 757 55 123 579 -31.29 +/- 11.89
ncm-et-9 05:55:15 1982655 2411 198 418 1795 -31.79 +/- 6.95
ncm-et-10 05:55:37 1986872 2411 236 400 1775 -23.67 +/- 7.09
ncm-et-11 01:49:00 2004003 736 77 126 533 -23.17 +/- 13.14
ncm-et-12 01:49:01 2017193 751 62 124 565 -28.75 +/- 12.28
ncm-et-13 05:54:41 1991771 2393 231 411 1751 -26.18 +/- 7.18
ncm-et-14 01:49:12 2020591 745 73 141 531 -31.8 +/- 13.29
ncm-et-15 05:55:34 1988624 2401 211 412 1778 -29.15 +/- 7.03
ncm-et-16 00:38:15 1988446 261 30 44 187 -18.65 +/- 22.43
20000 1825 3466 14709 -28.57 +/- 2.46

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
73664 ncm-et-15 2019-03-13 21:18 00:22:51 1979098 152 14 24 114 -22.89 +/- 27.55
73663 ncm-et-13 2019-03-13 21:18 00:22:49 1970326 150 14 30 106 -37.2 +/- 29.94
73662 ncm-et-9 2019-03-13 21:18 00:23:28 1965551 165 13 31 121 -38.05 +/- 27.14
73661 ncm-et-10 2019-03-13 21:15 00:25:50 1969224 167 17 28 122 -22.92 +/- 27.31
73660 ncm-et-4 2019-03-13 21:14 00:26:33 1965893 176 15 27 134 -23.73 +/- 24.99
73659 ncm-et-3 2019-03-13 21:13 00:27:55 1964457 190 16 25 149 -16.47 +/- 22.91
73658 ncm-et-15 2019-03-13 20:05 01:12:03 1973892 500 41 80 379 -27.16 +/- 14.89
73657 ncm-et-13 2019-03-13 20:03 01:13:42 1979221 500 68 86 346 -12.51 +/- 16.9
73656 ncm-et-9 2019-03-13 20:03 01:13:44 1961535 500 32 86 382 -37.67 +/- 14.6
73655 ncm-et-4 2019-03-13 20:01 01:12:21 1963997 500 36 97 367 -42.6 +/- 15.5
73654 ncm-et-3 2019-03-13 20:00 01:11:58 1977126 500 37 89 374 -36.26 +/- 15.13
73653 ncm-et-10 2019-03-13 20:00 01:14:17 1957858 500 48 89 363 -28.55 +/- 15.86
73652 ncm-et-9 2019-03-13 18:49 01:12:56 1966001 500 43 85 372 -29.25 +/- 15.31
73651 ncm-et-15 2019-03-13 18:48 01:16:38 1965863 500 48 93 359 -31.35 +/- 16.08
73650 ncm-et-3 2019-03-13 18:47 01:12:17 1965843 500 51 82 367 -21.57 +/- 15.66
73649 ncm-et-13 2019-03-13 18:47 01:15:05 1980681 500 41 86 373 -31.35 +/- 15.23
73648 ncm-et-4 2019-03-13 18:46 01:14:05 1972210 500 41 93 366 -36.26 +/- 15.62
73647 ncm-et-10 2019-03-13 18:46 01:13:19 1978979 500 39 95 366 -39.08 +/- 15.6
73646 ncm-et-13 2019-03-13 17:32 01:13:59 1969089 500 43 88 369 -31.35 +/- 15.48
73645 ncm-et-15 2019-03-13 17:32 01:14:55 1972192 500 39 84 377 -31.35 +/- 14.98
73644 ncm-et-4 2019-03-13 17:32 01:13:18 1969428 500 51 85 364 -23.66 +/- 15.83
73643 ncm-et-9 2019-03-13 17:31 01:16:24 1952355 500 40 78 382 -26.46 +/- 14.71
73642 ncm-et-3 2019-03-13 17:31 01:14:30 1963998 500 59 95 346 -25.06 +/- 16.86
73641 ncm-et-10 2019-03-13 17:31 01:13:12 1976757 500 43 83 374 -27.85 +/- 15.2
13493 ncm-et-11 2018-03-29 22:08 00:34:55 2014664 236 19 41 176 -32.48 +/- 22.19
13492 ncm-et-13 2018-03-29 22:08 00:35:14 2023936 243 25 42 176 -24.35 +/- 22.88
13491 ncm-et-15 2018-03-29 22:08 00:35:22 2018885 249 21 40 188 -26.56 +/- 21.25
13490 ncm-et-10 2018-03-29 22:08 00:35:46 2019533 244 28 31 185 -4.27 +/- 21.46
13489 ncm-et-3 2018-03-29 22:07 00:35:58 2025409 245 23 43 179 -28.43 +/- 22.48
13488 ncm-et-14 2018-03-29 22:07 00:36:15 2020999 245 25 58 162 -47.08 +/- 25.14
13487 ncm-et-12 2018-03-29 22:07 00:36:13 2013550 251 18 47 186 -40.32 +/- 21.62
13486 ncm-et-4 2018-03-29 22:07 00:36:20 1961794 254 23 35 196 -16.43 +/- 20.39
13485 ncm-et-9 2018-03-29 22:07 00:36:36 2025572 246 28 49 169 -29.73 +/- 24.22
13484 ncm-et-6 2018-03-29 22:07 00:36:39 2028197 254 25 49 180 -32.93 +/- 22.95
13483 ncm-et-5 2018-03-29 22:06 00:37:07 2022956 249 24 54 171 -42.06 +/- 23.97
13482 ncm-et-8 2018-03-29 22:06 00:37:07 2010427 257 13 40 204 -36.64 +/- 19.0
13481 ncm-et-16 2018-03-29 22:05 00:38:15 1988446 261 30 44 187 -18.65 +/- 22.43
13480 ncm-et-7 2018-03-29 22:04 00:39:28 2024427 266 39 47 180 -10.45 +/- 23.77
13479 ncm-et-9 2018-03-29 20:53 01:12:07 2024917 500 42 89 369 -32.76 +/- 15.47
13478 ncm-et-6 2018-03-29 20:53 01:12:13 2028197 500 47 102 351 -38.37 +/- 16.5
13477 ncm-et-10 2018-03-29 20:53 01:13:13 2018883 500 61 74 365 -9.04 +/- 15.83
13476 ncm-et-13 2018-03-29 20:53 01:13:52 2027374 500 40 79 381 -27.16 +/- 14.76
13475 ncm-et-4 2018-03-29 20:52 01:12:37 2022469 500 39 88 373 -34.16 +/- 15.21
13474 ncm-et-11 2018-03-29 20:52 01:14:05 1993342 500 58 85 357 -18.78 +/- 16.26
13473 ncm-et-12 2018-03-29 20:52 01:12:48 2020836 500 44 77 379 -22.96 +/- 14.92
13472 ncm-et-8 2018-03-29 20:52 01:11:47 2022468 500 42 83 375 -28.55 +/- 15.13
13471 ncm-et-14 2018-03-29 20:52 01:12:57 2020184 500 48 83 369 -24.36 +/- 15.53
13470 ncm-et-3 2018-03-29 20:52 01:13:09 2028031 500 46 95 359 -34.16 +/- 16.06
13469 ncm-et-15 2018-03-29 20:52 01:13:45 2021814 500 48 91 361 -29.95 +/- 15.97
13468 ncm-et-5 2018-03-29 20:52 01:12:08 2024590 500 49 80 371 -21.57 +/- 15.42
13467 ncm-et-7 2018-03-29 20:52 01:09:35 2022641 500 31 81 388 -34.86 +/- 14.24

Commit

Commit ID 613dc66c12b9bf357ae38129fb745dd8c2c035b1
Author Joona Kiiski
Date 2015-09-10 18:15:43 UTC
Careful SMP locking - Fix very occasional hangs Louis Zulli reported that Stockfish suffers from very occasional hangs with his 20 cores machine. Careful SMP debugging revealed that this was caused by "a ghost split point slave", where thread was marked as a split point slave, but wasn't actually working on it. The only logical explanation for this was double booking, where due to SMP race, the same thread is booked for two different split points simultaneously. Due to very intermittent nature of the problem, we can't say exactly how this happens. The current handling of Thread specific variables is risky though. Volatile variables are in some cases changed without spinlock being hold. In this case standard doesn't give us any kind of guarantees about how the updated values are propagated to other threads. We resolve the situation by enforcing very strict locking rules: - Values for key thread variables (splitPointsSize, activeSplitPoint, searching) can only be changed when the thread specific spinlock is held. - Structural changes for splitPoints[] are only allowed when the thread specific spinlock is held. - Thread booking decisions (per split point) can only be done when the thread specific spinlock is held. With these changes hangs didn't occur anymore during 2 days torture testing on Zulli's machine. We probably have a slight performance penalty in SMP mode due to more locking. STC (7 threads): ELO: -1.00 +-2.2 (95%) LOS: 18.4% Total: 30000 W: 4538 L: 4624 D: 20838 However stability is worth more than 1-2 ELO points in this case. No functional change Resolves #422
Copyright 2011–2024 Next Chess Move LLC