simple Makefile autodetect for ARCH and BITS #312
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
seylab
approved these changes
Jun 9, 2021
chussana-pin
approved these changes
Nov 11, 2021
dav1312
force-pushed
the
master
branch
2 times, most recently
from
9173bb1
to
e0afedd
Compare
Refine Evaluation Scaling with Piece-Specific Weights, instead of the simplified npm method. I took the initial idea from Viren6 , as he worked on it in September of last year. I worked on it, and tuned it, and now it passed both tests. Passed STC: LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 95712 W: 24731 L: 24325 D: 46656 Ptnml(0-2): 363, 11152, 24357, 11684, 300 https://tests.stockfishchess.org/tests/view/664b5493830eb9f886614af3 Passed LTC: LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 204480 W: 52167 L: 51501 D: 100812 Ptnml(0-2): 114, 22579, 56166, 23289, 92 https://tests.stockfishchess.org/tests/view/664b75dd830eb9f886614b44 closes #5277 Bench: 1384337
Passed non-regression STC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 65344 W: 16767 L: 16578 D: 31999 Ptnml(0-2): 198, 7557, 16987, 7718, 212 https://tests.stockfishchess.org/tests/view/664bd895830eb9f886615a26 Passed non-regression LTC: LLR: 2.96 (-2.94,2.94) <-1.75,0.25> Total: 35214 W: 8999 L: 8791 D: 17424 Ptnml(0-2): 16, 3804, 9760, 4010, 17 https://tests.stockfishchess.org/tests/view/664bead5830eb9f886615a52 closes #5278 Bench: 1296223
Params found after 30k spsa games at 60+0.6, with initial values from 64k spsa games at 45+0.45 First spsa with 64k / 120k games at 45+0.45: https://tests.stockfishchess.org/tests/view/664a561b5fc7b70b8817c663 https://tests.stockfishchess.org/tests/view/664ae88e830eb9f8866146f9 Second spsa with 30k / 120k games at 60+0.6: https://tests.stockfishchess.org/tests/view/664be227830eb9f886615a36 Values found at 10k games at 60+0.6 also passed STC and LTC: https://tests.stockfishchess.org/tests/view/664bf4bd830eb9f886615a72 https://tests.stockfishchess.org/tests/view/664c0905830eb9f886615abf Passed STC: https://tests.stockfishchess.org/tests/view/664c139e830eb9f886615af2 LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 69408 W: 18216 L: 17842 D: 33350 Ptnml(0-2): 257, 8275, 17401, 8379, 392 Passed LTC: https://tests.stockfishchess.org/tests/view/664cdaf7830eb9f886616a24 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 35466 W: 9075 L: 8758 D: 17633 Ptnml(0-2): 27, 3783, 9794, 4104, 25 closes #5280 bench 1301287
Also "fix" movepicker to allow depths between CHECKS and NO_CHECKS, which makes them easier to tweak (not that they get tweaked hardly ever) (This was more beneficial when there was a third stage to DEPTH_QS, but it's still an improvement now) closes #5205 No functional change
This patch is intended to prevent patches like 9b90cd8 and the subsequent reversion e3c9ed7 from happening again. Scaling behaviour of the reduction adjustments in the non-linear scaling section have been proven to >8 sigma: STC: https://tests.stockfishchess.org/tests/view/6647b19f6dcff0d1d6b05d52 Elo: 4.28 ± 0.8 (95%) LOS: 100.0% Total: 200000 W: 52555 L: 50094 D: 97351 Ptnml(0-2): 573, 22628, 51248, 24867, 684 nElo: 8.35 ± 1.5 (95%) PairsRatio: 1.10 VLTC: https://tests.stockfishchess.org/tests/view/6647b1b06dcff0d1d6b05d54 Elo: -1.48 ± 1.0 (95%) LOS: 0.2% Total: 100000 W: 25009 L: 25436 D: 49555 Ptnml(0-2): 11, 10716, 28971, 10293, 9 nElo: -3.23 ± 2.2 (95%) PairsRatio: 0.96 The else if condition is moved to the non scaling section based on: https://tests.stockfishchess.org/tests/view/664567a193ce6da3e93b3232 (It has no proven scaling) General comment improvements and removal of a redundant margin condition have also been included. closes #5266 No functional change
While the smallNet bool is no longer used as of now, setting it to false upon re-evaluation represents the correct eval state. closes #5279 No functional change
Passed Non-regression STC: LLR: 2.93 (-2.94,2.94) <-1.75,0.25> Total: 44896 W: 11600 L: 11388 D: 21908 Ptnml(0-2): 140, 5230, 11532, 5370, 176 https://tests.stockfishchess.org/tests/view/664cee31830eb9f886616a80 Passed Non-regression LTC: LLR: 2.96 (-2.94,2.94) <-1.75,0.25> Total: 56832 W: 14421 L: 14234 D: 28177 Ptnml(0-2): 37, 6251, 15643, 6458, 27 https://tests.stockfishchess.org/tests/view/664cfd4e830eb9f886616aa6 closes #5281 Bench: 1119412
This speedup was first inspired by a comment by @AndyGrant on my recent PR "If mullo_epi16 would preserve the signedness, then this could be used to remove 50% of the max operations during the halfkp-pairwise mat-mul relu deal." That got me thinking, because although mullo_epi16 did not preserve the signedness, mulhi_epi16 did, and so we could shift left and then use mulhi_epi16, instead of shifting right after the mullo. However, due to some issues with shifting into the sign bit, the FT weights and biases had to be multiplied by 2 for the optimisation to work. Speedup on "Arch=x86-64-bmi2 COMP=clang", courtesy of @Torom Result of 50 runs base (...es/stockfish) = 962946 +/- 1202 test (...ise-max-less) = 979696 +/- 1084 diff = +16750 +/- 1794 speedup = +0.0174 P(speedup > 0) = 1.0000 CPU: 4 x Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz Hyperthreading: on Also a speedup on "COMP=gcc", courtesy of Torom once again Result of 50 runs base (...tockfish_gcc) = 966033 +/- 1574 test (...max-less_gcc) = 983319 +/- 1513 diff = +17286 +/- 2515 speedup = +0.0179 P(speedup > 0) = 1.0000 CPU: 4 x Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz Hyperthreading: on Passed STC: LLR: 2.96 (-2.94,2.94) <0.00,2.00> Total: 67712 W: 17715 L: 17358 D: 32639 Ptnml(0-2): 225, 7472, 18140, 7759, 260 https://tests.stockfishchess.org/tests/view/664c1d75830eb9f886616906 closes #5282 No functional change
The patch regressed significantly at longer time controls. In particular, the `depth--` behavior was predicted to scale badly based on data from other variations of the patch. Passed VVLTC 1st sprt: https://tests.stockfishchess.org/tests/view/664d45cf830eb9f886616c7d LLR: 2.95 (-2.94,2.94) <0.00,2.00> Total: 51292 W: 13242 L: 12954 D: 25096 Ptnml(0-2): 5, 4724, 15896, 5020, 1 Passed VVLTC 2nd sprt: https://tests.stockfishchess.org/tests/view/664e641a928b1fb18de4e385 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 41884 W: 10933 L: 10634 D: 20317 Ptnml(0-2): 1, 3759, 13125, 4054, 3 closes #5283 Bench: 1503815
Passed non-regression STC: https://tests.stockfishchess.org/tests/view/664d033d830eb9f886616aff LLR: 2.93 (-2.94,2.94) <-1.75,0.25> Total: 102144 W: 26283 L: 26135 D: 49726 Ptnml(0-2): 292, 12201, 25991, 12243, 345 Passed non-regression LTC: https://tests.stockfishchess.org/tests/view/664d5c00830eb9f886616cb3 LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 250032 W: 63022 L: 63036 D: 123974 Ptnml(0-2): 103, 27941, 68970, 27871, 131 closes #5284 Bench: 1330940
Parameters were tuned in 2 stages: 1. 127k games at VVLTC: https://tests.stockfishchess.org/tests/view/6649f8dfb8fa20e74c39f52a. 2. 106k games at VVLTC: https://tests.stockfishchess.org/tests/view/664bfb77830eb9f886615a9d. Passed VVLTC 1st sprt: https://tests.stockfishchess.org/tests/view/664e8dd9928b1fb18de4e410 LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 20466 W: 5340 L: 5093 D: 10033 Ptnml(0-2): 0, 1796, 6397, 2037, 3 Passed VVLTC 2nd sprt: https://tests.stockfishchess.org/tests/view/664eb4aa928b1fb18de4e47d LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 15854 W: 4186 L: 3934 D: 7734 Ptnml(0-2): 1, 1367, 4938, 1621, 0 closes #5286 Bench: 1558110
This patch uses the same material weights for the nnue amplification term and the optimism term in evaluate(). STC: LLR: 2.99 (-2.94,2.94) <-1.75,0.25> Total: 83360 W: 21489 L: 21313 D: 40558 Ptnml(0-2): 303, 9934, 21056, 10058, 329 https://tests.stockfishchess.org/tests/view/664eee69928b1fb18de500d9 LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 192648 W: 48675 L: 48630 D: 95343 Ptnml(0-2): 82, 21484, 53161, 21501, 96 https://tests.stockfishchess.org/tests/view/664fa17aa86388d5e27d7d6e closes #5287 Bench: 1495602
The smallnet threshold is now below the training data range of the current smallnet (simple eval diff > 1k, nn-baff1edelf90.nnue) when no pawns are on the board. Params found with spsa at 93k / 120k games at 60+06: https://tests.stockfishchess.org/tests/view/664fa166a86388d5e27d7d6b Tuned on top of: #5287 Passed STC: https://tests.stockfishchess.org/tests/view/664fc8b7a86388d5e27d8dac LLR: 2.96 (-2.94,2.94) <0.00,2.00> Total: 64672 W: 16731 L: 16371 D: 31570 Ptnml(0-2): 239, 7463, 16517, 7933, 184 Passed LTC: https://tests.stockfishchess.org/tests/view/664fd5f9a86388d5e27d8dfe LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 210648 W: 53489 L: 52813 D: 104346 Ptnml(0-2): 102, 23129, 58164, 23849, 80 closes #5288 Bench: 1717838
This makes rootDelta logic easier to understand, recalculating the value where it belongs so removes an unnecessary branch. Passed non-regression STC: https://tests.stockfishchess.org/tests/view/664fc147a86388d5e27d8d8e LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 206016 W: 53120 L: 53089 D: 99807 Ptnml(0-2): 591, 20928, 59888, 21061, 540 closes #5289 No functional change
Passed Non-regression STC: LLR: 2.93 (-2.94,2.94) <-1.75,0.25> Total: 198432 W: 51161 L: 51119 D: 96152 Ptnml(0-2): 772, 23670, 50273, 23746, 755 https://tests.stockfishchess.org/tests/view/66517b9ea86388d5e27da966 Passed Non-regression LTC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 234150 W: 59200 L: 59197 D: 115753 Ptnml(0-2): 126, 26200, 64404, 26235, 110 https://tests.stockfishchess.org/tests/view/6653a84da86388d5e27daa63 closes #5292 bench 1555200
Tweak the return value formula in futility pruning. Passed STC: LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 60544 W: 15791 L: 15440 D: 29313 Ptnml(0-2): 193, 7024, 15520, 7309, 226 https://tests.stockfishchess.org/tests/view/6654ef22a86388d5e27db122 Passed LTC: LLR: 2.96 (-2.94,2.94) <0.50,2.50> Total: 126426 W: 32317 L: 31812 D: 62297 Ptnml(0-2): 55, 13871, 34869, 14350, 68 https://tests.stockfishchess.org/tests/view/66550644a86388d5e27db649 closes #5295 bench: 1856147
closes #5296 No functional change
Allow for NUMA memory replication for NNUE weights. Bind threads to ensure execution on a specific NUMA node. This patch introduces NUMA memory replication, currently only utilized for the NNUE weights. Along with it comes all machinery required to identify NUMA nodes and bind threads to specific processors/nodes. It also comes with small changes to Thread and ThreadPool to allow easier execution of custom functions on the designated thread. Old thread binding (WinProcGroup) machinery is removed because it's incompatible with this patch. Small changes to unrelated parts of the code were made to ensure correctness, like some classes being made unmovable, raw pointers replaced with unique_ptr. etc. Windows 7 and Windows 10 is partially supported. Windows 11 is fully supported. Linux is fully supported, with explicit exclusion of Android. No additional dependencies. ----------------- A new UCI option `NumaPolicy` is introduced. It can take the following values: ``` system - gathers NUMA node information from the system (lscpu or windows api), for each threads binds it to a single NUMA node none - assumes there is 1 NUMA node, never binds threads auto - this is the default value, depends on the number of set threads and NUMA nodes, will only enable binding on multinode systems and when the number of threads reaches a threshold (dependent on node size and count) [[custom]] - // ':'-separated numa nodes // ','-separated cpu indices // supports "first-last" range syntax for cpu indices, for example '0-15,32-47:16-31,48-63' ``` Setting `NumaPolicy` forces recreation of the threads in the ThreadPool, which in turn forces the recreation of the TT. The threads are distributed among NUMA nodes in a round-robin fashion based on fill percentage (i.e. it will strive to fill all NUMA nodes evenly). Threads are bound to NUMA nodes, not specific processors, because that's our only requirement and the OS can schedule them better. Special care is made that maximum memory usage on systems that do not require memory replication stays as previously, that is, unnecessary copies are avoided. On linux the process' processor affinity is respected. This means that if you for example use taskset to restrict Stockfish to a single NUMA node then the `system` and `auto` settings will only see a single NUMA node (more precisely, the processors included in the current affinity mask) and act accordingly. ----------------- We can't ensure that a memory allocation takes place on a given NUMA node without using libnuma on linux, or using appropriate custom allocators on windows (https://learn.microsoft.com/en-us/windows/win32/memory/allocating-memory-from-a-numa-node), so to avoid complications the current implementation relies on first-touch policy. Due to this we also rely on the memory allocator to give us a new chunk of untouched memory from the system. This appears to work reliably on linux, but results may vary. MacOS is not supported, because AFAIK it's not affected, and implementation would be problematic anyway. Windows is supported since Windows 7 (https://learn.microsoft.com/en-us/windows/win32/api/processtopologyapi/nf-processtopologyapi-setthreadgroupaffinity). Until Windows 11/Server 2022 NUMA nodes are split such that they cannot span processor groups. This is because before Windows 11/Server 2022 it's not possible to set thread affinity spanning processor groups. The splitting is done manually in some cases (required after Windows 10 Build 20348). Since Windows 11/Server 2022 we can set affinites spanning processor group so this splitting is not done, so the behaviour is pretty much like on linux. Linux is supported, **without** libnuma requirement. `lscpu` is expected. ----------------- Passed 60+1 @ 256t 16000MB hash: https://tests.stockfishchess.org/tests/view/6654e443a86388d5e27db0d8 ``` LLR: 2.95 (-2.94,2.94) <0.00,10.00> Total: 278 W: 110 L: 29 D: 139 Ptnml(0-2): 0, 1, 56, 82, 0 ``` Passed SMP STC: https://tests.stockfishchess.org/tests/view/6654fc74a86388d5e27db1cd ``` LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 67152 W: 17354 L: 17177 D: 32621 Ptnml(0-2): 64, 7428, 18408, 7619, 57 ``` Passed STC: https://tests.stockfishchess.org/tests/view/6654fb27a86388d5e27db15c ``` LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 131648 W: 34155 L: 34045 D: 63448 Ptnml(0-2): 426, 13878, 37096, 14008, 416 ``` fixes #5253 closes #5285 No functional change
Patch author: @ehsanrashid Passed STC: LLR: 2.93 (-2.94,2.94) <-1.75,0.25> Total: 116192 W: 30229 L: 30094 D: 55869 Ptnml(0-2): 451, 13880, 29351, 13911, 503 https://tests.stockfishchess.org/tests/view/66510a40a86388d5e27da936 Passed LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 441312 W: 111009 L: 111220 D: 219083 Ptnml(0-2): 217, 49390, 121659, 49167, 223 https://tests.stockfishchess.org/tests/view/66530696a86388d5e27da9e3 closes #5304 Bench: 1987574
This simplification patch merges the pawn count terms in the eval formula with the material term, updating the offset constant for the nnue part of the formula from 34000 to 34300 because the average pawn count in middlegame positions evaluated during search is around 8. STC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 138240 W: 35834 L: 35723 D: 66683 Ptnml(0-2): 527, 16587, 34817, 16626, 563 https://tests.stockfishchess.org/tests/view/6653f474a86388d5e27daaac LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 454272 W: 114787 L: 115012 D: 224473 Ptnml(0-2): 246, 51168, 124553, 50903, 266 https://tests.stockfishchess.org/tests/view/6654f256a86388d5e27db131 closes #5303 Bench: 1279635
Fractional bonus idea is from @Ergodice on [discord](https://discord.com/channels/435943710472011776/735707599353151579/1244039134499180614). Values are tuned for 149k games at LTC. SPSA tune: https://tests.stockfishchess.org/tests/view/6652d5d5a86388d5e27da9d6 Failed STC: LLR: -2.95 (-2.94,2.94) <0.00,2.00> Total: 67424 W: 17364 L: 17528 D: 32532 Ptnml(0-2): 238, 8043, 17299, 7909, 223 https://tests.stockfishchess.org/tests/view/66551e1ba86388d5e27db9f9 Passed LTC: LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 146910 W: 37141 L: 36695 D: 73074 Ptnml(0-2): 84, 16201, 40441, 16643, 86 https://tests.stockfishchess.org/tests/view/66559949a86388d5e27dcc5d Passed VLTC: LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 27248 W: 6924 L: 6633 D: 13691 Ptnml(0-2): 5, 2744, 7835, 3035, 5 https://tests.stockfishchess.org/tests/view/66563f4da86388d5e27dd27a closes #5299 Bench: 1390709
closes #5298 No functional change
Tested against PR #5299 Passed Non-regression STC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 76352 W: 19797 L: 19619 D: 36936 Ptnml(0-2): 236, 9017, 19509, 9161, 253 https://tests.stockfishchess.org/tests/view/66564f60a86388d5e27dd307 Passed Non-regression LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 114624 W: 28946 L: 28821 D: 56857 Ptnml(0-2): 59, 12675, 31714, 12810, 54 https://tests.stockfishchess.org/tests/view/6656543da86388d5e27dd329 closes #5301 Bench: 1212167
As stockfish nets and search evolve, the existing time control appears to give too little time at STC, roughly correct at LTC, and too little at VLTC+. This change adds an adjustment to the optExtra calculation. This adjustment is easy to retune and refine, so it should be easier to keep up-to-date than the more complex calculations used for optConstant and optScale. Passed STC 10+0.1: LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 169568 W: 43803 L: 43295 D: 82470 Ptnml(0-2): 485, 19679, 44055, 19973, 592 https://tests.stockfishchess.org/tests/view/66531865a86388d5e27da9fa Yellow LTC 60+0.6: LLR: -2.94 (-2.94,2.94) <0.50,2.50> Total: 209970 W: 53087 L: 52914 D: 103969 Ptnml(0-2): 91, 19652, 65314, 19849, 79 https://tests.stockfishchess.org/tests/view/6653e38ba86388d5e27daaa0 Passed VLTC 180+1.8 : LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 85618 W: 21735 L: 21342 D: 42541 Ptnml(0-2): 15, 8267, 25848, 8668, 11 https://tests.stockfishchess.org/tests/view/6655131da86388d5e27db95f closes #5297 Bench: 1212167
Passed non-regression VVLTC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 90792 W: 23155 L: 23018 D: 44619 Ptnml(0-2): 6, 8406, 28432, 8549, 3 https://tests.stockfishchess.org/tests/view/664ffa4ca86388d5e27d8e7a Passed non-regression VLTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 288136 W: 72608 L: 72659 D: 142869 Ptnml(0-2): 38, 30258, 83525, 30211, 36 https://tests.stockfishchess.org/tests/view/66551609a86388d5e27db9ae closes #5293 bench 1501735
#5608 STC: https://tests.stockfishchess.org/tests/view/66fb1bab86d5ee47d953b8cc LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 25536 W: 6797 L: 6560 D: 12179 Ptnml(0-2): 93, 2953, 6460, 3148, 114 LTC https://tests.stockfishchess.org/tests/view/66fb690e86d5ee47d953b8eb LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 225114 W: 57200 L: 57188 D: 110726 Ptnml(0-2): 197, 25076, 61995, 25096, 193 closes #5621 Bench: 1570076
Passed STC: LLR: 2.97 (-2.94,2.94) <-1.75,0.25> Total: 163680 W: 42689 L: 42605 D: 78386 Ptnml(0-2): 619, 19555, 41386, 19683, 597 https://tests.stockfishchess.org/tests/view/66f9451386d5ee47d953b7d9 Passed LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 96498 W: 24582 L: 24438 D: 47478 Ptnml(0-2): 62, 10642, 26718, 10744, 83 https://tests.stockfishchess.org/tests/view/66fd765786d5ee47d953ba1c closes #5622 Bench: 1309815
Reduce the size of the Magics table by half on modern cpu's and lay it out to match our access pattern. Namely we typically access the magics for the same square for both bishop and rook back to back so we want those to be in the same cache line. https://tests.stockfishchess.org/tests/view/6701c9b386d5ee47d953bcf4 LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 121664 W: 31931 L: 31497 D: 58236 Ptnml(0-2): 395, 13658, 32322, 14032, 425 A similar patch minus the size reduction finished yellow https://tests.stockfishchess.org/tests/view/6695f03f4ff211be9d4ec16c LLR: -2.94 (-2.94,2.94) <0.00,2.00> Total: 310688 W: 80940 L: 80746 D: 149002 Ptnml(0-2): 1119, 35032, 82846, 35230, 1117 closes #5623 No functional change
That allows 'make -j profile-build' work on ppc64 architectures, setting the use of the appropriate SIMD extension, Altivec or VSX. For VSX, gcc allows to map SSE2 intrinsics and get benefit of the existing SIMD code. On PowerMac G5, using altivec provides a performance improvement of 30%. On Talos 2, using vsx provides a performance improvement of 120%. closes #5624 No functional change
This patch introduces the value `meanSquaredScore`, which makes the initial delta sensitive to unstable iterative deepening scores. Passed STC: https://tests.stockfishchess.org/tests/view/66fed74286d5ee47d953bb42 LLR: 2.98 (-2.94,2.94) <0.00,2.00> Total: 71104 W: 18635 L: 18262 D: 34207 Ptnml(0-2): 234, 8365, 17993, 8714, 246 Passed LTC: https://tests.stockfishchess.org/tests/view/6700088e86d5ee47d953bbe9 LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 212544 W: 54238 L: 53560 D: 104746 Ptnml(0-2): 120, 23093, 59172, 23763, 124 closes #5627 Bench: 1395505
Passed STC: LLR: 2.98 (-2.94,2.94) <0.00,2.00> Total: 476160 W: 124285 L: 123311 D: 228564 Ptnml(0-2): 1662, 56266, 121219, 57302, 1631 https://tests.stockfishchess.org/tests/view/66ea3dc186d5ee47d953ae07 Failed Yellow LTC: LLR: -2.94 (-2.94,2.94) <0.50,2.50> Total: 230634 W: 58525 L: 58295 D: 113814 Ptnml(0-2): 113, 25301, 64299, 25451, 153 https://tests.stockfishchess.org/tests/view/66f1825e86d5ee47d953b2ec Passed Non-regression LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 112344 W: 28590 L: 28462 D: 55292 Ptnml(0-2): 71, 12439, 31039, 12537, 86 https://tests.stockfishchess.org/tests/view/6707474486d5ee47d953bfe3 closes #5629 Bench: 1283457
Size of low ply history should always be the same, so ensure it. closes #5630 No functional change
Created by setting output weights (256) and biases (8) of the previous main net nn-1111cefa1111.nnue to values found with spsa after 38k / 120k games at 120+1.2 using the same method as: #5459 nn-1111cefa1111.nnue -> nn-1cedc0ffeeee.nnue # weights changed: 185 mean: 0.0703 +/- 2.53 min: -6 max: 6 Passed STC: https://tests.stockfishchess.org/tests/view/6703589b86d5ee47d953bda1 LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 101984 W: 26690 L: 26275 D: 49019 Ptnml(0-2): 375, 11944, 25926, 12385, 362 Passed LTC: https://tests.stockfishchess.org/tests/view/670542d286d5ee47d953befa LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 106224 W: 27079 L: 26618 D: 52527 Ptnml(0-2): 71, 11508, 29487, 11981, 65 closes #5632 Bench: 1351413
Passed STC: LLR: 2.97 (-2.94,2.94) <-1.75,0.25> Total: 139360 W: 36143 L: 36033 D: 67184 Ptnml(0-2): 436, 16456, 35778, 16582, 428 https://tests.stockfishchess.org/tests/view/66fc49c786d5ee47d953b94e Passed LTC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 257748 W: 65163 L: 65184 D: 127401 Ptnml(0-2): 173, 28471, 71611, 28442, 177 https://tests.stockfishchess.org/tests/view/66ff01ae86d5ee47d953bb54 Passed LTC against rebased version: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 53610 W: 13691 L: 13501 D: 26418 Ptnml(0-2): 52, 5942, 14605, 6176, 30 https://tests.stockfishchess.org/tests/view/670a9c5c86d5ee47d953c231 closes #5633 Bench: 1282078
Passed STC: LLR: 2.92 (-2.94,2.94) <-1.75,0.25> Total: 138656 W: 36182 L: 36074 D: 66400 Ptnml(0-2): 523, 16422, 35310, 16570, 503 https://tests.stockfishchess.org/tests/view/6702beb386d5ee47d953bd41 Passed LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 680844 W: 172021 L: 172480 D: 336343 Ptnml(0-2): 492, 76259, 187419, 75720, 532 https://tests.stockfishchess.org/tests/view/67042b1f86d5ee47d953be7c closes #5634 Bench: 1169252
Passed STC: LLR: 2.96 (-2.94,2.94) <-1.75,0.25> Total: 80832 W: 21150 L: 20975 D: 38707 Ptnml(0-2): 283, 9531, 20598, 9736, 268 https://tests.stockfishchess.org/tests/view/670302fe86d5ee47d953bd68 Passed LTC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 46008 W: 11621 L: 11423 D: 22964 Ptnml(0-2): 30, 5072, 12606, 5262, 34 https://tests.stockfishchess.org/tests/view/6704074686d5ee47d953be53 Passed LTC Rebased: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 95814 W: 24340 L: 24195 D: 47279 Ptnml(0-2): 71, 10497, 26602, 10690, 47 https://tests.stockfishchess.org/tests/view/670ae1ac86d5ee47d953c262 closes #5636 Bench: 1119774
Passed STC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 45888 W: 12051 L: 11841 D: 21996 Ptnml(0-2): 123, 5356, 11807, 5504, 154 https://tests.stockfishchess.org/tests/view/670bb89086d5ee47d953c2d8 Passed LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 51336 W: 13021 L: 12830 D: 25485 Ptnml(0-2): 34, 5594, 14227, 5773, 40 https://tests.stockfishchess.org/tests/view/670c587f86d5ee47d953c31b closes #5637 Bench: 1192999
Passed STC: LLR: 2.97 (-2.94,2.94) <-1.75,0.25> Total: 92544 W: 23940 L: 23778 D: 44826 Ptnml(0-2): 286, 10936, 23638, 11154, 258 https://tests.stockfishchess.org/tests/view/670c3d6986d5ee47d953c30b Passed LTC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 43164 W: 10986 L: 10786 D: 21392 Ptnml(0-2): 27, 4713, 11905, 4907, 30 https://tests.stockfishchess.org/tests/view/670eda3d86d5ee47d953c51d closes #5639 Bench: 1281912
The Makefile is notoriously slow on windows, because of new processes being spawned I believe. This pr improves it a little bit for the help and config-sanity targets, with the latter also improving `make -j build` because it depends on that. On the same machine ubuntu (wsl) is more than 3 times faster, if there are other improvements we can make I'd be happy to hear about them. Ultimately #5543 also aims to improve this I believe, but it will take some additional time before that lands. ``` make config-sanity: patch: 6.199s master: 12.738s make help: patch: 3.1s master: 11.49s make -j build: patch: 36s master: 43.25s make -j build: master ubuntu: 10s ``` closes #5642 No functional change
After some simplifications bonuses and maluses are the same for quiet and non-quiet moves so it makes no sense to use quietMoveBonus/Malus, instead use just bonus/malus. closes #5649 No functional change
Avoids duplication of `using ... = Stats<int16_t, CORRECTION_HISTORY_LIMIT, COLOR_NB, CORRECTION_HISTORY_SIZE>;` closes #5650 No functional change Co-authored-by: Disservin <[email protected]>
Decreasing the number of operations Passed STC: LLR: 2.97 (-2.94,2.94) <-1.75,0.25> Total: 38880 W: 10038 L: 9823 D: 19019 Ptnml(0-2): 92, 4334, 10395, 4505, 114 https://tests.stockfishchess.org/tests/view/67112bf586d5ee47d953c6be Passed LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 242844 W: 61425 L: 61431 D: 119988 Ptnml(0-2): 145, 25175, 70797, 25151, 154 https://tests.stockfishchess.org/tests/view/6712387486d5ee47d953c737 closes #5655 Bench: 1281912
The definition of moveCountPruning may cause confusion by implying that the variable is unconstrained. However, once it is set to true, it should not be reset to false, otherwise it would break the internal logic of MovePicker. Several patches have overlooked this constraint. For example: https://tests.stockfishchess.org/tests/view/671e7c0486d5ee47d953d226 https://tests.stockfishchess.org/tests/view/66a1de7b4ff211be9d4eccea The implementation approach was suggested by Disservin. Passed non-regression STC: LLR: 3.02 (-2.94,2.94) <-1.75,0.25> Total: 180672 W: 47072 L: 47006 D: 86594 Ptnml(0-2): 536, 19482, 50247, 19522, 549 https://tests.stockfishchess.org/tests/view/6720df6f86d5ee47d953d542 closes #5661 No functional change
closes #5651 No functional change
This patch also allows improving flag to be true if static evaluation of the position is good enough. Passed STC: https://tests.stockfishchess.org/tests/view/6720906086d5ee47d953d4d0 LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 34816 W: 9172 L: 8858 D: 16786 Ptnml(0-2): 113, 3988, 8887, 4312, 108 Passed LTC: https://tests.stockfishchess.org/tests/view/6721162686d5ee47d953d597 LLR: 2.96 (-2.94,2.94) <0.50,2.50> Total: 145374 W: 37118 L: 36574 D: 71682 Ptnml(0-2): 91, 15875, 40212, 16417, 92 closes #5662 Bench: 1518856
https://tests.stockfishchess.org/tests/view/671bf61b86d5ee47d953cf23 And thanks to @xu-shawn for suggesting running a VLTC regress test since depth modifications affect scaling. Also, the LTC was showing a slight regress after 680+k games ~= -0.34 , for reference: https://tests.stockfishchess.org/tests/view/67042b1f86d5ee47d953be7c closes #5663 Bench: 1307308
closes #5666 No functional change
Tuning Run (90k Games): https://tests.stockfishchess.org/tests/view/67202b1c86d5ee47d953d442 Passed STC: LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 241024 W: 62616 L: 62001 D: 116407 Ptnml(0-2): 716, 28231, 62015, 28822, 728 https://tests.stockfishchess.org/tests/view/6725196786d5ee47d953d9f2 Passed LTC: LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 92532 W: 23678 L: 23246 D: 45608 Ptnml(0-2): 45, 9981, 25797, 10383, 60 https://tests.stockfishchess.org/tests/view/6727d3cb86d5ee47d953db9d closes #5667 Bench: 1066071
A single tuning run of 190k games was conducted: https://tests.stockfishchess.org/tests/view/670f3e3786d5ee47d953c554. Passed VVLTC 1st sprt: https://tests.stockfishchess.org/tests/view/672344dc86d5ee47d953d8c3 LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 56768 W: 14615 L: 14323 D: 27830 Ptnml(0-2): 3, 5152, 17789, 5430, 10 Passed VVLTC 2nd sprt (rebased): https://tests.stockfishchess.org/tests/view/6726d83786d5ee47d953db03 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 57884 W: 14885 L: 14554 D: 28445 Ptnml(0-2): 5, 5300, 17999, 5635, 3 closes #5669 Bench: 920336
At low enough depths, fail high with TT only when expected cutnode. Passed STC: https://tests.stockfishchess.org/tests/view/6726357b86d5ee47d953da8c LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 41184 W: 10873 L: 10551 D: 19760 Ptnml(0-2): 131, 4728, 10554, 5046, 133 Passed LTC: https://tests.stockfishchess.org/tests/view/6727326a86d5ee47d953db30 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 270888 W: 69040 L: 68243 D: 133605 Ptnml(0-2): 180, 29385, 75485, 30246, 148 closes #5670 Bench: 805776
Found by updating 489 L2 weights with values found from around 31k / 60k spsa games. Spsa was configured to use 60k games, down from 120k games in: #5459 623 spsa params: L2 weights from `nn-1cedc0ffeeee.nnue` where 24 <= |value| <= 30 A: 3000, alpha: 0.602, gamma: 0.101 weights: [-127, 127], c_end = 6 Passed STC: https://tests.stockfishchess.org/tests/view/6728d61e86d5ee47d953dcaf LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 187168 W: 48642 L: 48107 D: 90419 Ptnml(0-2): 558, 21888, 48213, 22311, 614 Passed LTC: https://tests.stockfishchess.org/tests/view/672b018f86d5ee47d953de98 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 235074 W: 59924 L: 59202 D: 115948 Ptnml(0-2): 131, 25467, 65610, 26207, 122 closes #5673 Bench: 898850
Passed STC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 37408 W: 9699 L: 9477 D: 18232 Ptnml(0-2): 130, 4326, 9577, 4534, 137 https://tests.stockfishchess.org/tests/view/672ffd8086d5ee47d953e633 Passed LTC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 151062 W: 38087 L: 37999 D: 74976 Ptnml(0-2): 63, 16686, 41958, 16748, 76 https://tests.stockfishchess.org/tests/view/673087aa86d5ee47d953e66b closes #5674 Bench: 848812
closes #5678 No functional change
Instead of using quiet histories use capture history with a different offset. Passed STC: https://tests.stockfishchess.org/tests/view/6731d5cc86d5ee47d953e719 LLR: 2.96 (-2.94,2.94) <0.00,2.00> Total: 428896 W: 111160 L: 110269 D: 207467 Ptnml(0-2): 1220, 50296, 110534, 51169, 1229 Passed LTC: https://tests.stockfishchess.org/tests/view/6733d9fd86d5ee47d953e962 LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 105882 W: 26918 L: 26458 D: 52506 Ptnml(0-2): 66, 11430, 29482, 11904, 59 closes #5679 Bench: 840721
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I added a simple autodetect flag for the makefile:
make build ARCH=autodetect### 2.1.1 Simple Defaults Autodetection ifeq ($(ARCH),autodetect) arch = $(shell uname -m) bits = $(shell getconf LONG_BIT) endifI used the modification in my docker build system.
Maybe its useful!