MateFinder 8

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

jhellis3
Posts: 546
Joined: Sat Aug 17, 2013 12:36 am

MateFinder 8

Post by jhellis3 »

I brought MateFinder into sync with the SF8 release.

No real changes other than that.

No compiles currently as profile-build is currently borked on my machine.

https://github.com/jhellis3/Stockfish/tree/mate_finder


As always, if you have positions which MF can not solve in a timely manner, please send them to me (mate in 16 or less), and I will try to track down the issue.
Vinvin
Posts: 5235
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: MateFinder 8

Post by Vinvin »

Thanks, Joseph !!
tpoppins
Posts: 919
Joined: Tue Nov 24, 2015 9:11 pm
Location: upstate

Re: MateFinder 8

Post by tpoppins »

Thank you for the update, Joseph.

Here are w32 and x64 Windoze bins (2.3 MB), compiled with MinGW:

TinyUpload
FileDropper

From the TXT:

Code: Select all

SF-MF-8-w32-old_tp.exe .... for old Win32 systems
SF-MF-8-w32-SSE_tp.exe .... for Win32 systems with SSE
SF-MF-8-x64_tp.exe ........ for old x64 systems
SF-MF-8-x64-POPCNT_tp.exe . for 'modern' x64 systems with POPCNT
SF-MF-8-x64-BMI2_tp.exe ... for Haswell (and newer) x64 systems

The w32 compiles have been built and tested on WinXP+SP3, the rest 
on Win7 x64 Pro. The BMI2 compile has not been tested and may not 
work.
Source link has also been included in the TXT.

edit: bench = 6993374
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: MateFinder 8

Post by MikeB »

Awesome Joe!

A MacOS Exe ( source , license etc included)
https://www.dropbox.com/s/82bg728bte7r5 ... S.zip?dl=1

Code: Select all

bench
===========================
Total time (ms) : 3784
Nodes searched  : 6993374
Nodes/second    : 1848143
[d]4k2r/p1q1np1p/1p2p1p1/2p1P1B1/6Q1/2P5/PP4PP/3R2K1 w k -
mate in 15

Code: Select all

setoption name Threads value 18
setoption name Hash value 2048
position fen 4k2r/p1q1np1p/1p2p1p1/2p1P1B1/6Q1/2P5/PP4PP/3R2K1 w k -
go movetime 180000
...
info depth 22 seldepth 43 multipv 1 score cp 2811 nodes 33360605 nps 19952514 hashfull 123 tbhits 0 time 1672 pv g4a4 c7c6 d1d8 e8d8 a4c6 f7f5 c6b7 h8e8 b7b8 d8d7 b8d6 d7c8 g5e7 c8b7 d6d7 b7a6 d7e8 f5f4 e8a4 a6b7 a4d7 b7a6 e7d6 f4f3 g2f3 g6g5 d7e6 g5g4 e6c8 a6b5 e5e6 b5c4 e6e7 c4d3 c8f5 d3d2 e7e8q d2c1 f3g4
info depth 23 seldepth 47 multipv 1 score mate 20 nodes 51746908 nps 22537851 hashfull 173 tbhits 0 time 2296 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5f6 g6g5 c6b7 h8e8 b7b8 d8d7 b8d6 d7c8 f6e7 e8e7 d6e7 g5g4 e7f7 g4g3 f7e6 c8c7 e6d6 c7b7 d6d7 b7a6 e5e6 g3h2 g1h2 h6h5 e6e7 a6a5 d7a7 a5b5 e7e8q b5c4 e8e4 c4b5 a7a4
...
info depth 47 seldepth 47 multipv 1 score mate 17 nodes 1228519447 nps 28818866 hashfull 1000 tbhits 0 time 42629 pv g4a4 c7c6 d1d8 e8d8 a4c6 c5c4 c6b7 h8e8 b7b8 d8d7 b8d6 d7c8 g5e7 e8e7 d6e7 g6g5 e7f7 b6b5 f7e6 c8b7 e6d7 b7b6 e5e6 g5g4 e6e7 g4g3 e7e8q g3h2 g1h2 a7a6 e8d8 b6c5 d7d5
info depth 48 currmove g4a4 currmovenumber 1
info depth 48 currmove g4f4 currmovenumber 2
info depth 48 currmove g5e3 currmovenumber 3
info depth 48 currmove g2g3 currmovenumber 4
info depth 48 currmove b2b3 currmovenumber 5
info depth 48 currmove g4e2 currmovenumber 6
info depth 48 currmove g5f6 currmovenumber 7
info depth 48 currmove g5f4 currmovenumber 8
info depth 48 currmove c3c4 currmovenumber 9
info depth 48 currmove d1d4 currmovenumber 10
info depth 48 currmove g5h6 currmovenumber 11
info depth 48 currmove a2a4 currmovenumber 12
info depth 48 currmove d1d7 currmovenumber 13
info depth 48 currmove g5d2 currmovenumber 14
info depth 48 currmove d1c1 currmovenumber 15
info depth 48 currmove h2h3 currmovenumber 16
info depth 48 currmove g4c4 currmovenumber 17
info depth 48 currmove g4f3 currmovenumber 18
info depth 48 currmove g1h1 currmovenumber 19
info depth 48 currmove g4e4 currmovenumber 20
info depth 48 currmove g4h5 currmovenumber 21
info depth 48 currmove g4f5 currmovenumber 22
info depth 48 currmove d1d8 currmovenumber 23
info depth 48 currmove g5c1 currmovenumber 24
info depth 48 currmove d1d5 currmovenumber 25
info depth 48 currmove d1a1 currmovenumber 26
info depth 48 currmove d1e1 currmovenumber 27
info depth 48 currmove g4d4 currmovenumber 28
info depth 48 currmove a2a3 currmovenumber 29
info depth 48 currmove d1d6 currmovenumber 30
info depth 48 currmove d1b1 currmovenumber 31
info depth 48 currmove d1f1 currmovenumber 32
info depth 48 currmove g5h4 currmovenumber 33
info depth 48 currmove b2b4 currmovenumber 34
info depth 48 currmove g4b4 currmovenumber 35
info depth 48 currmove g4h4 currmovenumber 36
info depth 48 currmove h2h4 currmovenumber 37
info depth 48 currmove g1f1 currmovenumber 38
info depth 48 currmove d1d2 currmovenumber 39
info depth 48 currmove g4g3 currmovenumber 40
info depth 48 currmove d1d3 currmovenumber 41
info depth 48 currmove g4h3 currmovenumber 42
info depth 48 currmove g1f2 currmovenumber 43
info depth 48 currmove g5e7 currmovenumber 44
info depth 48 currmove g4e6 currmovenumber 45
info depth 48 seldepth 47 multipv 1 score mate 15 nodes 4914933931 nps 31041165 hashfull 1000 tbhits 0 time 158336 pv g4a4 c7c6 d1d8 e8d8 a4c6 h8e8 c6d6 d8c8 g5e7 e8e7 d6e7 a7a5 e7f7 a5a4 f7e6 c8b7 e6d7 b7a6 d7a4 a6b7 e5e6 b6b5 a4b5 b7a7 e6e7 c5c4 e7e8q h7h5 e8b8
info depth 49 currmove g4a4 currmovenumber 1
info depth 49 seldepth 47 multipv 1 score mate 15 nodes 5613745400 nps 31187127 hashfull 1000 tbhits 0 time 180002 pv g4a4 c7c6 d1d8 e8d8 a4c6 h8e8 c6d6 d8c8 g5e7 e8e7 d6e7 a7a5 e7f7 a5a4 f7e6 c8b7 e6d7 b7a6 d7a4 a6b7 e5e6 b6b5 a4b5 b7a7 e6e7 c5c4 e7e8q h7h5 e8b8
bestmove g4a4 ponder c7c6
[d]r3r3/1ppp1ppk/1bn4p/1B2N3/3P4/2P4P/QP3PP1/6K1 w - -
mate-in-8

Code: Select all

...
info depth 9 seldepth 20 multipv 1 score cp 335 nodes 145858 nps 1585413 tbhits 0 time 92 pv b5d3 f7f5 d3f5 h7h8 e5g6 h8h7 a2f7 c6d4 c3d4 b6d4 f7d7 e8e1 g1h2
info depth 10 seldepth 20 multipv 1 score mate 8 nodes 387142 nps 1861259 tbhits 0 time 208 pv b5d3 h7g8 a2f7 g8h8 f7f5 a8a1 g1h2 a1h1 h2h1 g7g6 f5f6 h8h7 f6f7 h7h8 e5g6
now Stockfish-8-15y , once it finds mate - it simply stops - by design for faster play on the belief mate is mate - so different stockfishes for different folks - but I really like what Joe has done with matefinder

setoption name Hash value 2048
position fen r3r3/1ppp1ppk/1bn4p/1B2N3/3P4/2P4P/QP3PP1/6K1 w - -
go movetime 60000
info depth 1 seldepth 1 multipv 1 score cp 122 nodes 2315 nps 210454 tbhits 0 time 11 pv b5d3 h7h8 e5f7 h8g8
info depth 2 seldepth 4 multipv 1 score cp 71 nodes 4050 nps 311538 tbhits 0 time 13 pv a2b1 h7g8 e5d7
info depth 3 seldepth 5 multipv 1 score cp 93 nodes 5377 nps 358466 tbhits 0 time 15 pv a2b1 h7h8 e5d7 h8g8
info depth 4 seldepth 5 multipv 1 score cp 93 nodes 6561 nps 410062 tbhits 0 time 16 pv a2b1 h7h8 e5d7 h8g8 d7b6
info depth 5 seldepth 6 multipv 1 score cp 111 nodes 12621 nps 601000 tbhits 0 time 21 pv a2b1 h7g8 e5d7 b6a7 g2g3 g8h8
info depth 6 seldepth 7 multipv 1 score cp 85 nodes 16929 nps 705375 tbhits 0 time 24 pv a2b1 h7h8 e5d7 a8a5 b5d3 h8g8 d7b6
info depth 7 seldepth 8 multipv 1 score cp 106 nodes 33139 nps 946828 tbhits 0 time 35 pv a2b1 g7g6 e5d7 e8d8 d7b6 c7b6 b1d1 a8a5
info depth 8 seldepth 11 multipv 1 score cp 114 nodes 44327 nps 1030860 tbhits 0 time 43 pv a2b1 g7g6 e5d7 e8d8 d7b6 c7b6 b1d1 a8a5 b5d3
info depth 9 seldepth 17 multipv 1 score cp 303 nodes 110497 nps 1270080 tbhits 0 time 87 pv b5d3 h7h8 e5f7 h8g8 f7h6 g8h8 a2f7 c6d4 f7g6 a8a1 g1h2
info depth 10 seldepth 17 multipv 1 score cp 332 nodes 133971 nps 1353242 tbhits 0 time 99 pv b5d3 f7f5 d3f5 h7h8 e5g6 h8h7 a2f7 b6d4 g6f4 h7h8 c3d4 c6d4 f5d7 e8e1 g1h2
info depth 11 seldepth 20 multipv 1 score mate 8 nodes 297623 nps 1672039 tbhits 0 time 178 pv b5d3 h7g8 a2f7 g8h8 f7g6 h8g8 g6h7 g8f8 h7h8 f8e7 h8g7 e7e6 g7f7 e6d6 f7d7
bestmove b5d3 ponder h7g8


position fen 4k2r/p1q1np1p/1p2p1p1/2p1P1B1/6Q1/2P5/PP4PP/3R2K1 w k -
go movetime 180000
info depth 1 seldepth 1 multipv 1 score cp 153 nodes 1117 nps 372333 tbhits 0 time 3 pv g4f4
info depth 2 seldepth 2 multipv 1 score cp 110 nodes 2913 nps 728250 tbhits 0 time 4 pv g4a4 e8f8 g5h6 f8g8
info depth 3 seldepth 6 multipv 1 score cp 418 nodes 3737 nps 934250 tbhits 0 time 4 pv g4a4 c7c6 a4a7 e8g8
info depth 4 seldepth 6 multipv 1 score cp 418 nodes 4162 nps 832400 tbhits 0 time 5 pv g4a4 c7c6 a4a7 e8g8 g5e7
info depth 5 seldepth 10 multipv 1 score cp 476 nodes 6142 nps 1228400 tbhits 0 time 5 pv g4a4 c7c6 a4a7 e8g8 a7e7 f8a8
info depth 6 seldepth 10 multipv 1 score cp 476 nodes 6744 nps 1348800 tbhits 0 time 5 pv g4a4 c7c6 a4a7 e8g8 a7e7 f8a8 d1d8 a8d8
info depth 7 seldepth 10 multipv 1 score cp 476 nodes 8856 nps 1476000 tbhits 0 time 6 pv g4a4 c7c6 a4a7 e8g8 a7e7 f8a8 d1d8 a8d8 e7d8 g8g7
info depth 8 seldepth 13 multipv 1 score cp 906 nodes 23178 nps 1655571 tbhits 0 time 14 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 c6a8 d8c7 a8a7 c7c6 g5e7
info depth 9 seldepth 17 multipv 1 score cp 927 nodes 38072 nps 1903600 tbhits 0 time 20 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 c6a8 d8c7 a8a7 c7c6 g5e7 h8e8 e7f6
info depth 10 seldepth 17 multipv 1 score cp 1020 nodes 60154 nps 2148357 tbhits 0 time 28 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8e7 b8h8 h6h5 h8f6 e7e8
info depth 11 seldepth 21 multipv 1 score cp 1152 nodes 200653 nps 2605883 tbhits 0 time 77 pv g4a4 c7c6 d1d8 e8d8 a4c6 h8g8 g5e7 d8e7 c6d6 e7e8 d6b8 e8e7 b8g8 a7a5 g8a8 g6g5
info depth 12 seldepth 27 multipv 1 score cp 1220 nodes 280741 nps 2725640 tbhits 0 time 103 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8e7 b8h8 a7a5 h8b8 e7d7 b8b7 d7e8
info depth 13 seldepth 27 multipv 1 score cp 1253 nodes 381990 nps 2872105 tbhits 0 time 133 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8d7 b8h8 b6b5 h8a8 b5b4 a8a7 d7e8 a7b8 e8d7 c3b4
info depth 14 seldepth 28 multipv 1 score cp 1314 nodes 685853 nps 3250488 tbhits 0 time 211 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8e7 b8h8 b6b5 h8c8 c5c4 c8c5 e7e8 c5b5 e8f8 b5b8 f8g7 b8a7 g7f8 a7c5 f8g7 c5c4
info depth 15 seldepth 30 multipv 1 score cp 1325 nodes 858206 nps 3488642 tbhits 0 time 246 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8d7 b8h8 d7c6 h8a8 c6b5 a8a7 b5c4 a7f7 c4d5 f7g6 d5e5 g6g7 e5d6 g7h6
info depth 16 seldepth 30 multipv 1 score cp 1339 nodes 982719 nps 3666861 tbhits 0 time 268 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8d7 b8h8 d7c6 h8a8 c6b5 a8a7 b5c4 a7f7 c4d5 f7g6 d5e5 g6g7 e5d6
info depth 17 seldepth 30 multipv 1 score cp 1375 nodes 1544357 nps 4596300 tbhits 0 time 336 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8d7 b8h8 a7a5 h8a8 d7e7 a8b8 a5a4 b8b6 e7f8 b6c5 f8g7 c5a7 h6h5 a7a4
info depth 18 seldepth 30 multipv 1 score cp 1389 nodes 2016096 nps 5169476 tbhits 0 time 390 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8e7 b8h8 g6g5 h8b8 g5g4 b8a7 e7f8 a7b6 f8g7 b6c5 g4g3 h2g3 g7g6 c5e7
info depth 19 seldepth 32 multipv 1 score cp 1414 nodes 2441962 nps 5759344 tbhits 0 time 424 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8d7 b8h8 a7a5 h8a8 d7e7 a8b8 a5a4 b8b6 e7f8 b6c5 f8g7 c5a3 g6g5 a3e7 a4a3 e7a3 g7g6 a3f8
info depth 20 seldepth 37 multipv 1 score cp 1498 nodes 5969174 nps 10151656 tbhits 0 time 588 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5e7 d8e7 c6d6 e7e8 d6b8 e8e7 b8h8 b6b5 h8a8 a7a6 a8a6 b5b4 a6d6 e7e8 d6b8 e8d7 b8b7 d7d8 c3b4 c5b4 b7f7 h6h5 f7e6 g6g5
info depth 21 seldepth 40 multipv 1 score cp 2103 nodes 27562529 nps 21399479 hashfull 51 tbhits 0 time 1288 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5f6 g6g5 c6d6 d8c8 d6e7 h8h7 e7d8 c8b7 d8d7 b7a6 d7d3 a6b7 d3e4 b7c7 e4h7 c7c6 h7f7 c6d5 f7a7 d5e4 a7b6 h6h5 b6e6 g5g4 e6d7 h5h4 d7g4 e4e3 e5e6
info depth 22 seldepth 40 multipv 1 score cp 2823 nodes 42748728 nps 23922063 hashfull 89 tbhits 0 time 1787 pv g4a4 c7c6 d1d8 e8d8 a4c6 f7f5 c6b7 h8e8 b7b8 d8d7 b8d6 d7c8 g5e7 c8b7 d6d7 b7a6 d7a4 a6b7 a4e8 f5f4 e8d7 b7a6 d7e6 f4f3 g2f3 h7h5 e6g6 a6b7 e7d8 h5h4 e5e6 h4h3 e6e7 b7a6 e7e8q
info depth 23 seldepth 43 multipv 1 score cp 2989 nodes 50968817 nps 24646429 hashfull 109 tbhits 0 time 2068 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5f6 g6g5 c6d6 d8c8 d6e7 h8h7 e7d8 c8b7 d8d7 b7a6 d7d3 a6a5 d3h7 a5b5 h7f7 b5c4 f7e6 c4d3 e6d7 d3c2 e5e6 g5g4 e6e7 g4g3 d7a4 c2d2 h2g3 b6b5 a4a7 d2c2 e7e8q c2c1
info depth 24 seldepth 46 multipv 1 score cp 3114 nodes 57307601 nps 25156980 hashfull 129 tbhits 0 time 2278 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5f6 g6g5 c6d6 d8c8 d6e7 h8h7 e7d8 c8b7 d8d7 b7a6 d7d3 a6a5 d3h7 a5b5 h7f7 b5c4 f7e6 c4d3 e6d7 d3c2 e5e6 g5g4 e6e7 g4g3 d7a4 c2d2 h2g3 a7a5 e7e8q c5c4 a4c4 a5a4 c4a4 d2d3 a4d4 d3c2 d4b6
info depth 25 currmove g4f4 currmovenumber 2
info depth 25 currmove g5h6 currmovenumber 3
info depth 25 currmove d1d8 currmovenumber 4
info depth 25 currmove g5e3 currmovenumber 5
info depth 25 currmove b2b4 currmovenumber 6
info depth 25 currmove g5f6 currmovenumber 7
info depth 25 currmove g4d4 currmovenumber 8
info depth 25 currmove g4c4 currmovenumber 9
info depth 25 currmove g5f4 currmovenumber 10
info depth 25 currmove d1f1 currmovenumber 11
info depth 25 currmove g4f3 currmovenumber 12
info depth 25 currmove b2b3 currmovenumber 13
info depth 25 currmove d1d7 currmovenumber 14
info depth 25 currmove g4e2 currmovenumber 15
info depth 25 currmove a2a4 currmovenumber 16
info depth 25 currmove g4b4 currmovenumber 17
info depth 25 currmove g4f5 currmovenumber 18
info depth 25 currmove c3c4 currmovenumber 19
info depth 25 currmove g4e4 currmovenumber 20
info depth 25 currmove g4h4 currmovenumber 21
info depth 25 currmove g2g3 currmovenumber 22
info depth 25 currmove d1d6 currmovenumber 23
info depth 25 currmove g5h4 currmovenumber 24
info depth 25 currmove g5d2 currmovenumber 25
info depth 25 currmove d1d4 currmovenumber 26
info depth 25 currmove g5c1 currmovenumber 27
info depth 25 currmove g4h5 currmovenumber 28
info depth 25 currmove g4h3 currmovenumber 29
info depth 25 currmove d1d3 currmovenumber 30
info depth 25 currmove d1d2 currmovenumber 31
info depth 25 currmove d1e1 currmovenumber 32
info depth 25 currmove d1b1 currmovenumber 33
info depth 25 currmove d1c1 currmovenumber 34
info depth 25 currmove d1d5 currmovenumber 35
info depth 25 currmove a2a3 currmovenumber 36
info depth 25 currmove g4g3 currmovenumber 37
info depth 25 currmove h2h3 currmovenumber 38
info depth 25 currmove h2h4 currmovenumber 39
info depth 25 currmove g1h1 currmovenumber 40
info depth 25 currmove g1f2 currmovenumber 41
info depth 25 currmove d1a1 currmovenumber 42
info depth 25 currmove g1f1 currmovenumber 43
info depth 25 currmove g5e7 currmovenumber 44
info depth 25 currmove g4e6 currmovenumber 45
info depth 25 seldepth 46 multipv 1 score mate 21 nodes 88641790 nps 26967383 hashfull 192 tbhits 0 time 3287 pv g4a4 c7c6 d1d8 e8d8 a4c6 h7h6 g5f6 g6g5 c6d6 d8c8 d6e7 h8h7 e7d8 c8b7 d8d7 b7a6 d7d3 a6a5 d3h7 a5b5 h7f7 b5c4 f7e6 c4d3 e6d7 d3c2 e5e6 g5g4 e6e7 g4g3 e7e8q g3h2 g1h2 a7a5 e8e2 c2b1 d7d1 b1a2 d1a4 a2b1 e2c2
bestmove g4a4 ponder c7c6
Jouni
Posts: 3371
Joined: Wed Mar 08, 2006 8:15 pm

Re: MateFinder 8

Post by Jouni »

I tried Poppins compiles. BMI works, but it's 10% slower than popcount. But have there been any progress after SF MF110416 64 bmi2? I can't detect neither in mate solving nor in tactics. E.g. in Arasan suite with "study" flag SF MF110416 64 got 177, latest 162! Has there been any testing with some hundreds mates and combinations?
Jouni
Jimbo I
Posts: 149
Joined: Thu Feb 15, 2007 4:34 am
Location: USA

Re: MateFinder 8

Post by Jimbo I »

I couldn't find any information on the "clean search" option. Could someone please give a short, clear explanation of what this does?
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: MateFinder 8

Post by MikeB »

Jimbo I wrote:I couldn't find any information on the "clean search" option. Could someone please give a short, clear explanation of what this does?
It calls : Search::clear() which resets search state to zero, to obtain reproducible results., Search::clear() resets history, counter moves etc. You would not used this feature to play in a match - but you would use it to compare how the engine searches a move with "study" on or "study" off.
Jimbo I
Posts: 149
Joined: Thu Feb 15, 2007 4:34 am
Location: USA

Re: MateFinder 8

Post by Jimbo I »

Thanks Michael!
jhellis3
Posts: 546
Joined: Sat Aug 17, 2013 12:36 am

Re: MateFinder 8

Post by jhellis3 »

I tried Poppins compiles. BMI works, but it's 10% slower than popcount. But have there been any progress after SF MF110416 64 bmi2? I can't detect neither in mate solving nor in tactics. E.g. in Arasan suite with "study" flag SF MF110416 64 got 177, latest 162! Has there been any testing with some hundreds mates and combinations?
1) Obviously, if the BMI compile is slower something has gone horribly wrong, so it is a bit unfair to compare those two builds as that is potentially > 10% speed difference which is a massive handicap.

2) I would not suggest using the "study" option for any sort of test suites or as a measure of performance. It is a measure of last resort, only to be used for specific problems where everything else has failed.

3) There have not really been any huge changes to the actual MateFinder side of the code except one small improvement, and some better interactions with TBs resolving optimal mates. The vast majority of the code changes are simply the changes from Stockfish, which has shown to have made considerable progress over the last several months.

4) Has there been any testing with some hundreds mates and combinations? I test against many carefully selected positions. I have seen clear progress in terms of overall speed and a very small improvement in terms of number of positions solved. I always test with study disabled.

If you desire to perform an accurate comparison, I would at least ensure that the two builds are compiled identically. In terms of just general playing strength, of course, the latest MateFinder is significantly stronger (just as Stockfish is).
tpoppins
Posts: 919
Joined: Tue Nov 24, 2015 9:11 pm
Location: upstate

Re: MateFinder 8

Post by tpoppins »

jhellis3 wrote:
I tried Poppins compiles. BMI works, but it's 10% slower than popcount.
1) Obviously, if the BMI compile is slower something has gone horribly wrong.
It's a case of simple ignorance. Not having a Haswell box on hand, the clever provider of the above builds thought he'd work around it by commenting out the benchmarking section of the makefile. "Woohoo, it compiles now! And - oooh - look at the compilation time cut in half! Wonder what they put that section in for in the first place..."

Everyone's been very nice not to point fingers, thank you guys; but yes, that was a dumb thing to do.