Dev Builds » 20181127-0739

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 08:37:48 1964774 3330 1802 89 1439 +197.59 +/- 8.84
ncm-et-4 08:37:51 1961451 3339 1841 81 1417 +203.64 +/- 8.91
ncm-et-9 08:37:59 1969622 3353 1919 62 1372 +216.76 +/- 9.05
ncm-et-10 08:37:44 1964854 3322 1867 89 1366 +207.57 +/- 9.11
ncm-et-13 08:38:08 1974742 3350 1939 73 1338 +218.36 +/- 9.2
ncm-et-15 08:38:08 1953940 3306 1816 67 1423 +204.57 +/- 8.86
20000 11184 461 8355 +208.02 +/- 3.67

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
60997 ncm-et-15 2018-12-01 23:08 00:48:48 1904279 306 185 9 112 +227.64 +/- 32.21
60996 ncm-et-10 2018-12-01 23:06 00:49:53 1965071 322 187 8 127 +217.8 +/- 30.05
60995 ncm-et-3 2018-12-01 23:05 00:51:12 1964455 330 179 7 144 +200.82 +/- 27.92
60994 ncm-et-4 2018-12-01 23:04 00:52:35 1963843 339 186 8 145 +202.67 +/- 27.92
60993 ncm-et-13 2018-12-01 23:03 00:53:52 1978566 350 197 8 145 +209.91 +/- 27.98
60992 ncm-et-9 2018-12-01 23:02 00:54:58 1976748 353 198 4 151 +214.64 +/- 27.1
60991 ncm-et-15 2018-12-01 21:48 01:18:30 1965395 500 262 9 229 +193.64 +/- 21.94
60990 ncm-et-3 2018-12-01 21:47 01:16:48 1966771 500 269 11 220 +198.34 +/- 22.55
60989 ncm-et-4 2018-12-01 21:46 01:16:31 1969947 500 283 15 202 +207.95 +/- 23.81
60988 ncm-et-10 2018-12-01 21:45 01:20:30 1957401 500 291 10 199 +220.88 +/- 23.88
60987 ncm-et-13 2018-12-01 21:43 01:18:24 1977007 500 282 8 210 +213.85 +/- 23.08
60986 ncm-et-9 2018-12-01 21:43 01:17:21 1971572 500 293 8 199 +224.97 +/- 23.82
60985 ncm-et-15 2018-12-01 20:31 01:16:40 1962314 500 279 6 215 +212.86 +/- 22.68
60984 ncm-et-3 2018-12-01 20:29 01:17:02 1960615 500 271 13 216 +198.34 +/- 22.86
60983 ncm-et-4 2018-12-01 20:28 01:16:49 1971626 500 274 9 217 +205.04 +/- 22.67
60982 ncm-et-9 2018-12-01 20:25 01:16:42 1977475 500 282 10 208 +211.87 +/- 23.28
60981 ncm-et-10 2018-12-01 20:25 01:18:21 1969086 500 280 17 203 +203.11 +/- 23.78
60980 ncm-et-13 2018-12-01 20:24 01:17:48 1972084 500 304 10 186 +234.38 +/- 24.81
60979 ncm-et-15 2018-12-01 19:13 01:16:53 1962624 500 267 14 219 +193.64 +/- 22.7
60978 ncm-et-3 2018-12-01 19:11 01:17:06 1971600 500 271 18 211 +193.64 +/- 23.29
60977 ncm-et-4 2018-12-01 19:09 01:18:43 1961998 500 278 7 215 +210.89 +/- 22.72
60976 ncm-et-10 2018-12-01 19:08 01:16:08 1965847 500 281 14 205 +206.98 +/- 23.59
60975 ncm-et-13 2018-12-01 19:07 01:16:04 1966613 500 288 11 201 +216.85 +/- 23.78
60974 ncm-et-9 2018-12-01 19:07 01:17:16 1966305 500 286 13 201 +212.86 +/- 23.83
60973 ncm-et-15 2018-12-01 17:53 01:18:52 1965079 500 269 9 222 +200.24 +/- 22.36
60972 ncm-et-3 2018-12-01 17:51 01:18:58 1962302 500 281 8 211 +212.86 +/- 23.02
60971 ncm-et-13 2018-12-01 17:50 01:15:48 1975985 500 293 9 198 +223.94 +/- 23.92
60970 ncm-et-4 2018-12-01 17:50 01:17:25 1962302 500 273 14 213 +199.29 +/- 23.07
60969 ncm-et-10 2018-12-01 17:50 01:17:02 1974876 500 274 14 212 +200.24 +/- 23.14
60968 ncm-et-9 2018-12-01 17:49 01:17:28 1965380 500 273 16 211 +197.4 +/- 23.25
60967 ncm-et-15 2018-12-01 16:32 01:19:16 1961085 500 273 10 217 +203.11 +/- 22.71
60966 ncm-et-3 2018-12-01 16:32 01:17:44 1963532 500 269 14 217 +195.51 +/- 22.82
60965 ncm-et-13 2018-12-01 16:32 01:17:22 1976397 500 282 10 208 +211.87 +/- 23.28
60964 ncm-et-10 2018-12-01 16:31 01:18:02 1959242 500 269 11 220 +198.34 +/- 22.55
60963 ncm-et-9 2018-12-01 16:31 01:16:59 1960775 500 288 6 206 +221.9 +/- 23.27
60962 ncm-et-4 2018-12-01 16:30 01:18:56 1936284 500 290 13 197 +216.85 +/- 24.1
60961 ncm-et-3 2018-12-01 15:12 01:18:58 1964148 500 262 18 220 +185.33 +/- 22.73
60960 ncm-et-9 2018-12-01 15:12 01:17:15 1969103 500 299 5 196 +234.38 +/- 23.91
60959 ncm-et-13 2018-12-01 15:12 01:18:50 1976543 500 293 17 190 +215.85 +/- 24.67
60958 ncm-et-4 2018-12-01 15:12 01:16:52 1964157 500 257 15 228 +183.51 +/- 22.19
60957 ncm-et-15 2018-12-01 15:12 01:19:09 1956805 500 281 10 209 +210.89 +/- 23.22
60956 ncm-et-10 2018-12-01 15:12 01:17:48 1962461 500 285 15 200 +209.91 +/- 23.94

Commit

Commit ID de7182f4eee33029a7b79f6b097db06357669127
Author Steinar H. Gunderson
Date 2018-11-27 07:39:23 UTC
Turn on MADV_RANDOM for Syzygy mmaps (on Unix-like builds) When running on a cloud VM (n1-highcpu-96) with several NVMe SSDs and some non-SSDs for tablebases, I noticed that the average SSD request size was more than 256 kB. This doesn't make a lot of sense for Syzygy tablebases, which have a block size of 32 bytes and very low locality. Seemingly, the tablebase access patterns during probing make the OS, at least Linux, think that readahead is advantageous; normally, it gives up doing readahead if there are too many misses, but it doesn't, perhaps due to the fairly high overall hit rates. (It seems the kernel cannot distinguish between reading a block that was paged in because the userspace wanted it explicitly, and one that was read as part of readahead.) Setting MADV_RANDOM effectively turns off readahead, which causes the request size to drop to 4 kB. In the aforemented cloud VM test, this roughly tripled the amount of I/O requests that were able to go through, while reducing the total traffic from 2.8 GB/sec to 56 MB/sec (moving the bottleneck to the non-SSDs; it seems the SSDs could have sustained many more requests). Closes https://github.com/official-stockfish/Stockfish/pull/1829 No functional change.
Copyright 2011–2024 Next Chess Move LLC