Dev Builds » 20210730-1515

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 08:17:09 1965418 3351 2769 5 577 +407.1 ± 14.16
ncm-et-4 08:17:20 1959360 3353 2838 4 511 +430.53 ± 15.06
ncm-et-9 08:17:14 1960906 3341 2855 10 476 +438.37 ± 15.63
ncm-et-10 08:17:36 1953219 3305 2777 9 519 +421.37 ± 14.95
ncm-et-13 08:17:08 1957093 3317 2816 7 494 +432.53 ± 15.33
ncm-et-15 08:17:17 1959275 3333 2816 8 509 +427.23 ± 15.1
20000 16871 43 3086 +425.94 ± 6.12

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
150476 ncm-et-10 2021-07-31 01:15 00:46:29 1949030 305 245 0 60 +384.88 ± 44.39
150475 ncm-et-13 2021-07-31 01:12 00:48:36 1956792 317 267 1 49 +423.24 ± 49.62
150474 ncm-et-15 2021-07-31 01:11 00:49:48 1961398 333 281 3 49 +418.27 ± 49.64
150473 ncm-et-4 2021-07-31 01:09 00:52:01 1971070 353 297 0 56 +425.88 ± 46.16
150472 ncm-et-9 2021-07-31 01:08 00:52:45 1956943 341 290 0 51 +436.98 ± 48.5
150471 ncm-et-3 2021-07-31 01:08 00:52:53 1958038 351 291 0 60 +411.75 ± 44.5
150470 ncm-et-10 2021-07-30 23:59 01:15:30 1955420 500 429 3 68 +438.95 ± 41.92
150469 ncm-et-13 2021-07-30 23:57 01:14:01 1959237 500 424 0 76 +433.94 ± 39.45
150468 ncm-et-15 2021-07-30 23:57 01:13:02 1963696 500 408 1 91 +395.65 ± 35.96
150467 ncm-et-4 2021-07-30 23:54 01:14:26 1961419 500 416 1 83 +412.8 ± 37.74
150466 ncm-et-3 2021-07-30 23:54 01:13:23 1973489 500 402 2 96 +381.7 ± 35.02
150465 ncm-et-9 2021-07-30 23:53 01:14:36 1958169 500 409 4 87 +391.57 ± 36.91
150464 ncm-et-10 2021-07-30 22:42 01:15:49 1956787 500 424 1 75 +431.48 ± 39.8
150463 ncm-et-13 2021-07-30 22:42 01:15:12 1952101 500 419 1 80 +419.61 ± 38.47
150462 ncm-et-15 2021-07-30 22:40 01:16:20 1957719 500 422 1 77 +426.65 ± 39.25
150461 ncm-et-3 2021-07-30 22:39 01:14:35 1958780 500 418 0 82 +419.61 ± 37.91
150460 ncm-et-9 2021-07-30 22:38 01:13:48 1964165 500 429 1 70 +444.09 ± 41.26
150459 ncm-et-4 2021-07-30 22:38 01:15:17 1938805 500 423 0 77 +431.48 ± 39.18
150458 ncm-et-13 2021-07-30 21:26 01:14:46 1950321 500 431 2 67 +446.7 ± 42.25
150457 ncm-et-10 2021-07-30 21:26 01:15:14 1946359 500 428 1 71 +441.5 ± 40.95
150456 ncm-et-9 2021-07-30 21:25 01:13:14 1964489 500 436 1 63 +463.16 ± 43.6
150455 ncm-et-15 2021-07-30 21:25 01:15:18 1957894 500 422 3 75 +421.93 ± 39.85
150454 ncm-et-3 2021-07-30 21:24 01:13:48 1971255 500 422 0 78 +429.05 ± 38.92
150453 ncm-et-4 2021-07-30 21:23 01:14:07 1963867 500 424 1 75 +431.48 ± 39.8
150452 ncm-et-13 2021-07-30 20:12 01:13:46 1967090 500 426 0 74 +438.95 ± 40.01
150451 ncm-et-10 2021-07-30 20:11 01:14:23 1958965 500 420 0 80 +424.27 ± 38.4
150450 ncm-et-4 2021-07-30 20:10 01:12:30 1961848 500 416 2 82 +410.58 ± 38.02
150449 ncm-et-15 2021-07-30 20:10 01:13:53 1962768 500 431 0 69 +452.04 ± 41.5
150448 ncm-et-3 2021-07-30 20:10 01:14:00 1964203 500 418 0 82 +419.61 ± 37.91
150447 ncm-et-9 2021-07-30 20:09 01:15:26 1943135 500 441 0 59 +481.09 ± 45.07
150446 ncm-et-10 2021-07-30 18:56 01:14:24 1954063 500 418 2 80 +415.05 ± 38.52
150445 ncm-et-3 2021-07-30 18:55 01:13:45 1968622 500 409 0 91 +399.81 ± 35.89
150444 ncm-et-13 2021-07-30 18:55 01:16:26 1952578 500 422 1 77 +426.65 ± 39.25
150443 ncm-et-15 2021-07-30 18:54 01:14:53 1952521 500 427 0 73 +441.5 ± 40.29
150442 ncm-et-9 2021-07-30 18:54 01:13:52 1967235 500 426 3 71 +431.48 ± 41.0
150441 ncm-et-4 2021-07-30 18:54 01:15:36 1950813 500 444 0 56 +490.71 ± 46.32
150440 ncm-et-9 2021-07-30 17:40 01:13:33 1972209 500 424 1 75 +431.48 ± 39.8
150439 ncm-et-15 2021-07-30 17:40 01:14:03 1958934 500 425 0 75 +436.43 ± 39.73
150438 ncm-et-13 2021-07-30 17:40 01:14:21 1961536 500 427 2 71 +436.43 ± 40.99
150437 ncm-et-4 2021-07-30 17:40 01:13:23 1967701 500 418 0 82 +419.61 ± 37.91
150436 ncm-et-3 2021-07-30 17:40 01:14:45 1963543 500 409 3 88 +393.6 ± 36.68
150435 ncm-et-10 2021-07-30 17:40 01:15:47 1951912 500 413 2 85 +404.05 ± 37.32

Commit

Commit ID 26edf9534ad571a6d26bf9db47d21776cbf45d54
Author Tomasz Sobczyk
Date 2021-07-30 15:15:52 UTC
Avoid unnecessary stores in the affine transform This patch improves the codegen in the AffineTransform::forward function for architectures >=SSSE3. Current code works directly on memory and the compiler cannot see that the stores through outptr do not alias the loads through weights and input32. The solution implemented is to perform the affine transform with local variables as accumulators and only store the result to memory at the end. The number of accumulators required is OutputDimensions / OutputSimdWidth, which means that for the 1024->16 affine transform it requires 4 registers with SSSE3, 2 with AVX2, 1 with AVX512. It also cuts the number of stores required by NumRegs * 256 for each node evaluated. The local accumulators are expected to be assigned to registers, but even if this cannot be done in some case due to register pressure it will help the compiler to see that there is no aliasing between the loads and stores and may still result in better codegen. See https://godbolt.org/z/59aTKbbYc for codegen comparison. passed STC: LLR: 2.94 (-2.94,2.94) <-0.50,2.50> Total: 140328 W: 10635 L: 10358 D: 119335 Ptnml(0-2): 302, 8339, 52636, 8554, 333 closes https://github.com/official-stockfish/Stockfish/pull/3634 No functional change
Copyright 2011–2024 Next Chess Move LLC