Dev Builds » 20211207-1708

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 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 11:10:10 1960329 3328 2827 4 497 +434.26 ± 15.27
ncm-et-4 11:09:53 1958808 3332 2879 5 448 +452.78 ± 16.1
ncm-et-9 11:10:03 1961467 3351 2862 6 483 +439.31 ± 15.5
ncm-et-10 11:10:19 1957047 3310 2827 3 480 +440.44 ± 15.54
ncm-et-13 11:10:13 1958661 3335 2853 6 476 +441.08 ± 15.62
ncm-et-15 11:10:10 1963861 3344 2863 4 477 +442.74 ± 15.59
20000 17111 28 2861 +441.7 ± 6.36

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
154951 ncm-et-10 2021-12-08 18:07 01:03:43 1960154 310 258 1 51 +411.72 ± 48.56
154950 ncm-et-3 2021-12-08 18:05 01:05:02 1978254 328 287 1 40 +465.97 ± 55.32
154949 ncm-et-13 2021-12-08 18:03 01:07:01 1962152 335 275 0 60 +402.87 ± 44.46
154948 ncm-et-4 2021-12-08 18:02 01:08:12 1954052 332 294 1 37 +481.93 ± 57.68
154947 ncm-et-15 2021-12-08 18:01 01:09:38 1968468 344 297 1 46 +449.98 ± 51.35
154946 ncm-et-9 2021-12-08 18:00 01:10:01 1964324 351 295 0 56 +424.81 ± 46.16
154945 ncm-et-10 2021-12-08 16:26 01:40:06 1954808 500 420 0 80 +424.27 ± 38.4
154944 ncm-et-3 2021-12-08 16:24 01:40:22 1939195 500 417 1 82 +415.05 ± 37.98
154943 ncm-et-13 2021-12-08 16:22 01:40:51 1956378 500 421 1 78 +424.28 ± 38.99
154942 ncm-et-4 2021-12-08 16:21 01:39:56 1960921 500 426 0 74 +438.95 ± 40.01
154941 ncm-et-15 2021-12-08 16:21 01:38:59 1963688 500 418 0 82 +419.61 ± 37.91
154940 ncm-et-9 2021-12-08 16:20 01:39:58 1964918 500 429 0 71 +446.7 ± 40.89
154939 ncm-et-10 2021-12-08 14:44 01:41:48 1953892 500 429 0 71 +446.7 ± 40.89
154938 ncm-et-3 2021-12-08 14:43 01:40:33 1957862 500 434 0 66 +460.32 ± 42.48
154937 ncm-et-4 2021-12-08 14:41 01:39:59 1957704 500 432 0 68 +454.76 ± 41.82
154936 ncm-et-13 2021-12-08 14:40 01:41:10 1957404 500 436 1 63 +463.16 ± 43.6
154935 ncm-et-9 2021-12-08 14:40 01:39:09 1961695 500 427 0 73 +441.5 ± 40.29
154934 ncm-et-15 2021-12-08 14:38 01:42:49 1951154 500 441 2 57 +474.93 ± 45.97
154933 ncm-et-3 2021-12-08 13:04 01:38:51 1967699 500 426 0 74 +438.95 ± 40.01
154932 ncm-et-10 2021-12-08 13:01 01:41:42 1957248 500 426 0 74 +438.95 ± 40.01
154931 ncm-et-13 2021-12-08 12:59 01:40:18 1957868 500 433 0 67 +457.52 ± 42.15
154930 ncm-et-15 2021-12-08 12:59 01:38:01 1968010 500 422 1 77 +426.65 ± 39.25
154929 ncm-et-9 2021-12-08 12:58 01:41:48 1955902 500 432 2 66 +449.35 ± 42.58
154928 ncm-et-4 2021-12-08 12:57 01:42:41 1934594 500 433 0 67 +457.52 ± 42.15
154927 ncm-et-3 2021-12-08 11:21 01:42:08 1958326 500 421 0 79 +426.65 ± 38.66
154926 ncm-et-10 2021-12-08 11:20 01:40:52 1961540 500 432 1 67 +452.04 ± 42.22
154925 ncm-et-15 2021-12-08 11:18 01:40:17 1960461 500 441 0 59 +481.09 ± 45.07
154924 ncm-et-13 2021-12-08 11:17 01:41:09 1955588 500 437 1 62 +466.04 ± 43.97
154923 ncm-et-4 2021-12-08 11:17 01:40:11 1966961 500 432 0 68 +454.76 ± 41.82
154922 ncm-et-9 2021-12-08 11:15 01:41:23 1953738 500 425 2 73 +431.48 ± 40.4
154921 ncm-et-3 2021-12-08 09:38 01:42:12 1958475 500 413 1 86 +406.2 ± 37.04
154920 ncm-et-10 2021-12-08 09:38 01:41:12 1948771 500 435 1 64 +460.32 ± 43.24
154919 ncm-et-13 2021-12-08 09:37 01:39:43 1960921 500 423 1 76 +429.05 ± 39.52
154918 ncm-et-15 2021-12-08 09:37 01:40:42 1970331 500 424 0 76 +433.94 ± 39.45
154917 ncm-et-4 2021-12-08 09:36 01:40:03 1970804 500 436 1 63 +463.16 ± 43.6
154916 ncm-et-9 2021-12-08 09:36 01:39:18 1962307 500 423 1 76 +429.05 ± 39.52
154915 ncm-et-4 2021-12-08 07:56 01:38:51 1966624 500 426 3 71 +431.48 ± 41.0
154914 ncm-et-9 2021-12-08 07:56 01:38:26 1967388 500 431 1 68 +449.35 ± 41.89
154913 ncm-et-13 2021-12-08 07:56 01:40:01 1960316 500 428 2 70 +438.95 ± 41.29
154912 ncm-et-3 2021-12-08 07:56 01:41:02 1962494 500 429 1 70 +444.09 ± 41.26
154911 ncm-et-10 2021-12-08 07:56 01:40:56 1962921 500 427 0 73 +441.5 ± 40.29
154910 ncm-et-15 2021-12-08 07:56 01:39:44 1964919 500 420 0 80 +424.27 ± 38.4

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 2011–2024 Next Chess Move LLC