This server will shut down on March 30th. You can download development versions of Stockfish here https://github.com/official-stockfish/Stockfish/releases
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Andreas Matthies
Date: Mon Nov 17 13:37:55 2025 +0100 Timestamp: 1763383075 Cleanup benchmark Avoid unnecessary consecutive ucinewgame commands in benchmark setup. Remove measuring time and nodes in speedtest warmup. closes https://github.com/official-stockfish/Stockfish/pull/6427 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Joost VandeVondele
Date: Mon Nov 17 13:36:11 2025 +0100 Timestamp: 1763382971 Silence warning for systems without shared memory support Fixes https://github.com/official-stockfish/Stockfish/issues/6425 closes https://github.com/official-stockfish/Stockfish/pull/6426 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Carlos Esparza
Date: Mon Nov 17 13:34:45 2025 +0100 Timestamp: 1763382885 inline make_index() and avoid templating perspective combination of two patches. pass perspective as argument passed STC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 52832 W: 13725 L: 13528 D: 25579 Elo +1.30 Ptnml(0-2): 154, 5756, 14412, 5927, 167 https://tests.stockfishchess.org/tests/view/69162e307ca8781852331c6a inline make_index passed STC: LLR: 2.93 (-2.94,2.94) <-1.75,0.25> Total: 68768 W: 17786 L: 17607 D: 33375 Elo +0.90 Ptnml(0-2): 187, 7591, 18694, 7680, 232 https://tests.stockfishchess.org/tests/view/6916859e7ca8781852331d36 closes https://github.com/official-stockfish/Stockfish/pull/6429 No functional change see source |
| Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Timothy Herchen
Date: Mon Nov 17 13:24:32 2025 +0100 Timestamp: 1763382272 Use non-locking operations for nodes count and tbHits fetch_add and friends must produce a locking instruction because they guarantee consistency in the presence of multiple writers. Because only one thread ever writes to nodes and tbHits, we may use relaxed load and store to emit regular loads and stores (on both x86-64 and arm64), while avoiding undefined behavior and miscounting nodes. Credit must go to Arseniy Surkov of Reckless (codedeliveryservice) for uncovering this performance gotcha. failed yellow as a gainer on fishtest: LLR: -2.95 (-2.94,2.94) <0.00,2.00> Total: 219616 W: 57165 L: 57105 D: 105346 Elo +0.09 Ptnml(0-2): 732, 24277, 59720, 24357, 722 https://tests.stockfishchess.org/tests/view/69086abfea4b268f1fac2809 potential small speedup on large core system: ``` ==== master ==== 1 Nodes/second : 294269736 2 Nodes/second : 295217629 Average (over 2): 294743682 ==== patch ==== 1 Nodes/second : 299003603 2 Nodes/second : 298111320 Average (over 2): 298557461 ``` closes https://github.com/official-stockfish/Stockfish/pull/6392 No functional change see source |
| Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Shawn Xu
Date: Fri Nov 14 17:43:44 2025 +0100 Timestamp: 1763138624 Fix undefined behavior C++ standard does not define `uint8_t` as an allowed pointer alias type. Despite `uint8_t` being `unsigned char` on most platforms, this is not enforced by the standard. https://eel.is/c++draft/basic.lval#11 https://stackoverflow.com/a/16138470 closes https://github.com/official-stockfish/Stockfish/pull/6420 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Robert Nurnberg
Date: Fri Nov 14 17:38:55 2025 +0100 Timestamp: 1763138335 Allow time checking after each DTZ probe This PR allows for time checking within Tablebases::rank_root_moves(). The principal application for now is syzygy_extend_pv(). In the past, Stockfish suffered some time losses at e.g. TCEC when the HDD with the DTZ tables was too slow for the chosen move overhead setting, see #5894. ``` > ./fastchess -engine name="patch" cmd=./stockfish.patch -engine name="master" cmd=./stockfish.master -each tc=10+0.1 option.SyzygyPath=/disk1/syzygy/3-4-5-6/WDL:/disk2/syzygy/3-4-5-6/DTZ -draw movenumber=34 movecount=8 score=20 -openings file=UHO_Lichess_4852_v1.epd format=epd -rounds 50 -repeat -concurrency 16 ... Results of patch vs master (10+0.1, 1t, 16MB, UHO_Lichess_4852_v1.epd): Elo: 59.64 +/- 35.35, nElo: 117.61 +/- 68.10 LOS: 99.96 %, DrawRatio: 56.00 %, PairsRatio: 4.50 Games: 100, Wins: 34, Losses: 17, Draws: 49, Points: 58.5 (58.50 %) Ptnml(0-2): [0, 4, 28, 15, 3], WL/DD Ratio: 0.87 -------------------------------------------------- Player: master Timeouts: 19 Crashed: 0 Finished match ``` closes https://github.com/official-stockfish/Stockfish/pull/6422 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Shawn Xu
Date: Fri Nov 14 17:32:17 2025 +0100 Timestamp: 1763137937 Pass accumulator caches by reference closes https://github.com/official-stockfish/Stockfish/pull/6416 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Daniel Monroe
Date: Fri Nov 14 17:30:55 2025 +0100 Timestamp: 1763137855 Clean up code and comments closes https://github.com/official-stockfish/Stockfish/pull/6416 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 + SSSE3 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Shawn Xu
Date: Fri Nov 14 17:29:01 2025 +0100 Timestamp: 1763137741 Cleanup, fix style issues use naming convention for variables, functions, constants closes https://github.com/official-stockfish/Stockfish/pull/6416 no functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: AliceRoselia
Date: Fri Nov 14 17:28:17 2025 +0100 Timestamp: 1763137697 Simplify indexing Removes 1 function (from_to) from the Move type, change them to simpler raw. Remove the "and" use in the correction history structure calculation. Adjust the indexing. Passed simplification STC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 42880 W: 11327 L: 11113 D: 20440 Elo +1.73 Ptnml(0-2): 161, 4852, 11212, 5042, 173 https://tests.stockfishchess.org/tests/view/690593b8ea4b268f1fac1e8a Passed simplification LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 43560 W: 11276 L: 11079 D: 21205 Elo +1.57 Ptnml(0-2): 17, 4679, 12197, 4864, 23 https://tests.stockfishchess.org/tests/view/6906f819ea4b268f1fac216b closes https://github.com/official-stockfish/Stockfish/pull/6391 Bench: 2523092 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: FauziAkram
Date: Fri Nov 14 17:24:16 2025 +0100 Timestamp: 1763137456 Simplify NMP reduction formula Passed STC: LLR: 2.97 (-2.94,2.94) <-1.75,0.25> Total: 178912 W: 46625 L: 46559 D: 85728 Elo +0.13 Ptnml(0-2): 540, 21167, 45975, 21235, 539 https://tests.stockfishchess.org/tests/view/69060677ea4b268f1fac1f25 Passed LTC: LLR: 2.99 (-2.94,2.94) <-1.75,0.25> Total: 140988 W: 36278 L: 36176 D: 68534 Elo +0.25 Ptnml(0-2): 82, 15520, 39215, 15568, 109 https://tests.stockfishchess.org/tests/view/6908bc32ea4b268f1fac288f closes https://github.com/official-stockfish/Stockfish/pull/6416 bench: 2959834 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Timothy Herchen
Date: Thu Nov 13 22:42:10 2025 +0100 Timestamp: 1763070130 check for material key validity in tbprobe During recent work on threat inputs, there was a hard-to-debug crash in tbprobe caused by incorrectly computing st->materialKey. This PR adds correctness checking to pos_is_ok and a faster check in tbprobe that will actually run in CI (because pos_is_ok checking is disabled even in debug mode, for performance purposes). closes https://github.com/official-stockfish/Stockfish/pull/6410 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Torsten Hellwig
Date: Thu Nov 13 22:36:30 2025 +0100 Timestamp: 1763069790 Print NEON before POPCNT in compilation settings Normally, Stockfish outputs the compilation settings from advanced to less advanced, e.g.: `AVX512ICL VNNI AVX512 BMI2 AVX2 SSE41 SSSE3 SSE2 POPCNT` With NEON, however, POPCNT is printed first, followed by the more advanced NEON options: `POPCNT NEON_DOTPROD` This PR places POPCNT behind the NEON options as well. closes https://github.com/official-stockfish/Stockfish/pull/6402 no functional change see source |
| Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Daniel Monroe
Date: Thu Nov 13 22:35:33 2025 +0100 Timestamp: 1763069733 Simplify threat term in movepick Passed simplification STC LLR: 2.97 (-2.94,2.94) <-1.75,0.25> Total: 71296 W: 18605 L: 18419 D: 34272 Elo +0.91 Ptnml(0-2): 250, 8374, 18183, 8622, 219 https://tests.stockfishchess.org/tests/view/690454c7ea4b268f1fac1bbe Passed simplification LTC LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 240552 W: 61870 L: 61874 D: 116808 Elo -0.01 Ptnml(0-2): 113, 26300, 67460, 26284, 119 https://tests.stockfishchess.org/tests/view/69063956ea4b268f1fac1f66 closes https://github.com/official-stockfish/Stockfish/pull/6401 bench 2697501 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Didier Durand
Date: Thu Nov 13 22:32:12 2025 +0100 Timestamp: 1763069532 Fix some doc typos closes https://github.com/official-stockfish/Stockfish/pull/6400 No functional change see source |
| Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Daniel Monroe
Date: Thu Nov 13 22:31:02 2025 +0100 Timestamp: 1763069462 Remove check term in capture movepick Passed simplification STC LLR: 2.96 (-2.94,2.94) <-1.75,0.25> Total: 248448 W: 64697 L: 64708 D: 119043 Elo -0.02 Ptnml(0-2): 784, 29393, 63971, 29202, 874 https://tests.stockfishchess.org/tests/view/68fc7afb637acd2a11e72d86 Passed simplification LTC LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 193626 W: 49808 L: 49764 D: 94054 Elo +0.08 Ptnml(0-2): 162, 21415, 53621, 21447, 168 https://tests.stockfishchess.org/tests/view/6901ad09637acd2a11e73828 closes https://github.com/official-stockfish/Stockfish/pull/6395 bench 2920273 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Guenther Demetz
Date: Thu Nov 13 22:26:35 2025 +0100 Timestamp: 1763069195 Unify do_move & do_null_move workload While being there also remove and unused assignment. Tested for non-regression at STC: https://tests.stockfishchess.org/tests/view/69060d81ea4b268f1fac1f36 LLR: 2.93 (-2.94,2.94) <-1.75,0.25> Total: 94496 W: 24570 L: 24421 D: 45505 Elo +0.55 Ptnml(0-2): 264, 10145, 26275, 10306, 258 closes https://github.com/official-stockfish/Stockfish/pull/6383 no functional change see source |
| Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: mstembera
Date: Thu Nov 13 22:21:35 2025 +0100 Timestamp: 1763068895 Fix MSVC compile broken after Shared Memory patch. closes https://github.com/official-stockfish/Stockfish/pull/6397 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Timothy Herchen
Date: Thu Nov 13 22:17:03 2025 +0100 Timestamp: 1763068623 Don't copy around DirtyThreats The contents of DirtyThreats gets memcpyed twice for each call to do_move. (Return value optimization doesn't apply to the do_move function itself because it constructs a std::pair, so it gets copied; and the calls to reset also require a copy.) This patch inserts the dirty info in place. Sometimes the caller of do_move ignores the DirtyThreats info, so we pass in scratch objects. I found that stack-allocating these scratch objects was bad on Fishtest, so I begrudgingly put them in the Position struct. Both templating the do_move function on whether dirty threats are needed and putting a null-check branch for each use of dirty threats were slowdowns locally. Of course, nothing prevents a future attempt at cleaning this up. passed STC LLR: 2.96 (-2.94,2.94) <0.00,2.00> Total: 68448 W: 17770 L: 17418 D: 33260 Elo +1.79 Ptnml(0-2): 198, 7425, 18630, 7769, 202 https://tests.stockfishchess.org/tests/view/6914c01a7ca87818523318ba closes https://github.com/official-stockfish/Stockfish/pull/6414 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Viren6
Date: Thu Nov 13 22:13:47 2025 +0100 Timestamp: 1763068427 Improve Threats Speed Passed STC: LLR: 2.95 (-2.94,2.94) <0.00,2.00> Total: 23168 W: 6132 L: 5845 D: 11191 Elo +4.30 Ptnml(0-2): 77, 2405, 6325, 2708, 69 https://tests.stockfishchess.org/tests/view/69148c3c7ca8781852331831 ``` Result of 50 runs ================== base (./stockfish.master ) = 985641 +/- 4249 test (./stockfish.patch ) = 1038567 +/- 5679 diff = +52926 +/- 4473 speedup = +0.0537 P(speedup > 0) = 1.0000 CPU: 16 x AMD Ryzen 9 3950X 16-Core Processor Hyperthreading: on ``` closes https://github.com/official-stockfish/Stockfish/pull/6415 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Daniel Monroe
Date: Thu Nov 13 22:10:30 2025 +0100 Timestamp: 1763068230 Refine constant in correction history update Passed STC LLR: 3.04 (-2.94,2.94) <0.00,2.00> Total: 250112 W: 65277 L: 64635 D: 120200 Elo +0.89 Ptnml(0-2): 841, 29326, 64134, 29860, 895 https://tests.stockfishchess.org/tests/view/69096c46ea4b268f1fac2a39 Passed LTC LLR: 2.96 (-2.94,2.94) <0.50,2.50> Total: 142908 W: 37141 L: 36604 D: 69163 Elo +1.31 Ptnml(0-2): 100, 15478, 39742, 16053, 81 https://tests.stockfishchess.org/tests/view/690e7dfbec1d00d2c195c351 closes https://github.com/official-stockfish/Stockfish/pull/6405 bench 2613869 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: sscg13
Date: Wed Nov 12 10:49:39 2025 +0100 Timestamp: 1762940979 Update NNUE architecture to SFNNv10 with Threat Inputs and net nn-49c1193b131c.nnue This commit introduces Full Threat Input features, which are a subset of Piece(Square)-Piece(Square) pairs. In any given position, the active features consist of pairs where the second piece’s square lies in the attack set of the first piece. This is an extremely simplified explanation that leaves out many details. The already-used HalfKAv2_hm feature set completes the input features. Minor quantization changes have also been made. The net nn-49c1193b131c.nnue was trained by vondele using the following setup: https://github.com/vondele/nettest/blob/7de71238e9b295e3f88ed7c9c5936af632c9b981/threats.yaml A graphical version of an earlier scheme (with less refinement) that illustrates the core concepts can be found attached. [NewInputs.pdf](https://github.com/user-attachments/files/23478441/NewInputs.pdf) Further information, as well as a brief description of the history of development, can be found attached. [Stockfish threat inputs PR summary.pdf](https://github.com/user-attachments/files/23478634/Stockfish.threat.inputs.PR.summary.pdf) This has been a huge effort spanning over half a year, with the original [discussion thread](https://discord.com/channels/435943710472011776/1336647760388034610) reaching over 11k messages. Thanks to everyone who has contributed. Monty PRs: https://github.com/official-monty/Monty/pull/87 (Initial threat input PR) https://github.com/official-monty/Monty/pull/114 (Fixed threat indexing to take into account colour correctly) https://github.com/official-monty/Monty/pull/116 (i8 quantisation of weights whilst keeping calculations in i16) Yukari commit: https://github.com/yukarichess/yukari/commit/2d482c64a79cec03cf4987d5289334b9cdc737bc (Threat inputs merged) Plentychess PRs: https://github.com/Yoshie2000/PlentyChess/pull/400 (Threat inputs merged) https://github.com/Yoshie2000/PlentyChess/pull/411 (Threat input weights quantised to i8) Passed STC: LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 63424 W: 16956 L: 16591 D: 29877 Elo +2.00 Ptnml(0-2): 276, 7522, 15797, 7795, 322 https://tests.stockfishchess.org/tests/view/69105b3dec1d00d2c195c569 Passed LTC: LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 27876 W: 7417 L: 7110 D: 13349 Elo +3.83 Ptnml(0-2): 23, 3033, 7530, 3318, 34 https://tests.stockfishchess.org/tests/view/6910d817ec1d00d2c195c66e Passed VVLTC (Hash accidentally set to 1/2 normal value for both sides): LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 12458 W: 3353 L: 3102 D: 6003 Elo +7.00 Ptnml(0-2): 0, 1106, 3767, 1355, 1 https://tests.stockfishchess.org/tests/view/69115a26ec1d00d2c195c7cd This version has also passed non-regression LTC against the originally passed version: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 51144 W: 13086 L: 12903 D: 25155 Elo +1.24 Ptnml(0-2): 22, 5167, 15018, 5336, 29 https://tests.stockfishchess.org/tests/view/69138a317ca87818523314bf LTC elo estimate on ARM: 1 patch : 13.9 1.9 38296.5 73728 52 2 master : 0.0 ---- 35431.5 73728 48 closes https://github.com/official-stockfish/Stockfish/pull/6411 bench: 2626086 Co-authored-by: Shawn Xu <> Co-authored-by: Timothy Herchen <> Co-authored-by: Viren6 <94880762+> Co-authored-by: Yoshie2000 <> Co-authored-by: Joost Vandevondele <> Co-authored-by: rn5f107s2 <> Co-authored-by: cj5716 <125858804+> Co-authored-by: AliceRoselia <63040919+> Co-authored-by: Linmiao Xu <> Co-authored-by: Disservin <> see source |
| Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Tomasz Sobczyk
Date: Sun Nov 2 16:04:09 2025 +0100 Timestamp: 1762095849 Use shared memory for network weights This enables different Stockfish processes that use the same weights to use the same memory. The approach establishes equivalence by memory content, and is compatible with NUMA replication. The benefit of sharing is reduced memory usage and a speedup thanks to improved (inter-process) caching of the network in the CPUs cache, and thus reduced bandwidth usage to main memory. Even though this change doesn't benefit a user running a single process, this helps on fishtest or e.g. for Lichess, when multiple games run concurrently, or multiple positions are analyzed in parallel. This concept was probably first introduced in the Monty engine (https://github.com/official-monty/Monty/pull/62), after a discussion in https://github.com/official-stockfish/fishtest/issues/2077 on the issue of memory pressure. Measurements based on Torch (https://github.com/user-attachments/files/21386224/verbatim.pdf) further suggested that large gains were possible. Multiple other engines have adopted this 'verbatim' format as well. The implementation here adds the flexibility needed for SF, for example, retains the ability to bundle compressed networks with the binary, to load nets by uci option, and to distribute the shared nets to the proper NUMA region. This flexibility comes with a fair amount of complexity in the implementation, such as OS specific code, and fallback code. For most users this should be transparent. However, for example, those running docker containers should ensure the `--ipc` flag is set correctly, and `--shm-size` is sufficiently large. The benefits of this patch significantly depend on hardware, with systems with many cores and a large (O(150MB), the net size) L3 cache benefitting typically most. On such systems SF speedups (as measured via nps playing games with large concurrency but just 1 thread) can be 38%, which results in master vs. patch Elo which gains about 25 Elo. ``` # PLAYER : RATING ERROR POINTS PLAYED (%) 1 shared_memoryPR : 24.8 1.9 39432.0 73728 53 2 master : 0.0 ---- 34296.0 73728 47 ``` In a multithreaded setup, where weights are already shared, that benefit is smaller, for example on the same HW as above, but with 8t for each side. ``` # PLAYER : RATING ERROR POINTS PLAYED (%) 1 shared_memoryPR : 5.2 3.5 9351.0 18432 51 2 master : 0.0 ---- 9081.0 18432 49 ``` On fishtest with a typical hardware mix of our contributors, the following was measured: STC, 60k games https://tests.stockfishchess.org/tests/view/69074a49ea4b268f1fac236c Elo: 4.69 ± 1.4 (95%) LOS: 100.0% Total: 60000 W: 16085 L: 15275 D: 28640 Elo +4.69 Ptnml(0-2): 154, 6440, 16053, 7148, 205 nElo: 9.38 ± 2.8 (95%) PairsRatio: 1.12 To verify correctness with a single process on a NUMA architecture, speedtest was used, confirming near equivalence: ``` master: Average (over 10): 296236186 shared_memory: Average (over 10): 295769332 ``` Currently, using large pages for the shared network weights is not always possible, which can lead to a small slowdown (1-2%), in case a single process is run. closes https://github.com/official-stockfish/Stockfish/pull/6173 No functional change Co-authored-by: disservin <> Co-authored-by: Joost VandeVondele <> see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: FauziAkram
Date: Sat Nov 1 11:39:00 2025 +0100 Timestamp: 1761993540 Simplify Futility pruning formula Passed STC: LLR: 2.93 (-2.94,2.94) <-1.75,0.25> Total: 170112 W: 44266 L: 44193 D: 81653 Elo +0.15 Ptnml(0-2): 599, 20062, 43611, 20235, 549 https://tests.stockfishchess.org/tests/view/68f50b94637acd2a11e721ad Passed LTC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 164658 W: 42393 L: 42318 D: 79947 Elo +0.16 Ptnml(0-2): 120, 18127, 45747, 18228, 107 https://tests.stockfishchess.org/tests/view/68fa6683637acd2a11e72ac3 closes https://github.com/official-stockfish/Stockfish/pull/6389 bench: 2351426 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 + SSSE3 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Shawn Xu
Date: Sat Nov 1 11:36:33 2025 +0100 Timestamp: 1761993393 Simplify Corrhist Bonus Passed Non-regression STC: LLR: 2.96 (-2.94,2.94) <-1.75,0.25> Total: 187776 W: 48687 L: 48631 D: 90458 Elo +0.10 Ptnml(0-2): 570, 22327, 48090, 22279, 622 https://tests.stockfishchess.org/tests/view/68f58019637acd2a11e72233 Passed Non-regression LTC: LLR: 2.96 (-2.94,2.94) <-1.75,0.25> Total: 159378 W: 41027 L: 40946 D: 77405 Elo +0.18 Ptnml(0-2): 122, 17645, 44078, 17718, 126 https://tests.stockfishchess.org/tests/view/68fe7492637acd2a11e73090 closes https://github.com/official-stockfish/Stockfish/pull/6385 Bench: 2458738 see source |