Ponder may contain a bug

tissatussa commented 1 year ago

hi, i used your Linux asset binarie(s) to run in CuteChess, but at some point in a game Booot fails by "connection lost" .. it happened several times .. i tried to solve it by setting Ponder OFF (*) and that seems to help !? Is it a bug ?

*) in CuteChess settings, Pondering is called "Thinking in opponent's time"

booot76 commented 1 year ago

Booot does not use Ponder mode by default. Please, send parameters do you use in cutechess-cli and parameters of you machine.

tissatussa commented 1 year ago

Booot does not use Ponder mode by default

OK, but the option in CuteChess should work, am i right ?

..send parameters do you use in cutechess-cli..

i only used the GUI : in the game settings window i set the checkbox "Thinking in opponent's time" to TRUE.

..and parameters of you machine..

$ uname -a
Linux roelof-HP-Elite-x2-1012-G2 5.15.0-67-generic #74-Ubuntu SMP Wed Feb 22 14:14:39 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux

is this information good for you ?

booot76 commented 1 year ago

Thanks. How much memory does you machine have? Seems the problem is in my data structure. They are unoptimal and may cause a problems. I will look on it tonight and most probably will change the binaries.

tissatussa commented 1 year ago

How much memory does you machine have?

8 Gb RAM, SSD

btw. very few engines cause memory problems when using CuteChess, but Booot is NOT one of them.

tissatussa commented 1 year ago

you might create an UCI option to write a log file ? The normal UCI info output lines do not show any error etc. -- the engine just stops ..

tissatussa commented 1 year ago

i managed to compile the current Booot 7.1 source (see my #21) : now i have a valid AVX2-PEXT binary ! With this version i did another quick test : a quick engine-vs-engine game from a certain starting position, with Ponder ON .. the same happens : Booot 7.1 suddenly stops .. here's the screenshot and the (last) log lines :

Booot-7-1-connection-stalls

[...]

<Booot v7.1(8): info currmovenumber 19 info currmove b6b5
<Booot v7.1(8): info currmovenumber 20 info currmove g6g5
<Booot v7.1(8): info currmovenumber 21 info currmove h6h5
<Booot v7.1(8): info depth 29 time 13402 nodes 17382099 nps 1296977
<Booot v7.1(8): bestmove c4e5 ponder g4g3
>SF Fat Titz Hybrid(7): ponderhit
>Booot v7.1(8): position fen 8/p5pp/1pk5/5p2/P1nn4/2NN3P/5PPK/8 w - - 0 1 moves h2g3 a7a5 g3f4 h7h6 h3h4 g7g6 g2g4 d4e6 f4g3 f5g4 g3g4 e6c5 d3f4 c4e5 g4g3
>Booot v7.1(8): isready
<Booot v7.1(8): readyok
>Booot v7.1(8): go ponder wtime 329888 btime 260767 winc 3000 binc 3000
<SF Fat Titz Hybrid(7): info depth 28 currmove g4h3 currmovenumber 2
<SF Fat Titz Hybrid(7): info depth 29 seldepth 51 multipv 1 score cp -436 upperbound nodes 4446934 nps 821376 hashfull 317 tbhits 0 time 5414 pv g4g3 c5d3
<SF Fat Titz Hybrid(7): info depth 29 currmove g4g3 currmovenumber 1
<SF Fat Titz Hybrid(7): info depth 29 seldepth 51 multipv 1 score cp -420 lowerbound nodes 4464940 nps 820911 hashfull 318 tbhits 0 time 5439 pv g4g3

[...]

<SF Fat Titz Hybrid(7): info depth 40 currmove f4h3 currmovenumber 19
<SF Fat Titz Hybrid(7): info depth 40 currmove f4g6 currmovenumber 20
<SF Fat Titz Hybrid(7): info depth 40 seldepth 51 multipv 1 score cp -432 upperbound nodes 177140488 nps 688202 hashfull 1000 tbhits 0 time 257396 pv f4e6 e5d7
<SF Fat Titz Hybrid(7): info depth 40 currmove f4e6 currmovenumber 1
>Booot v7.1(8): stop
>SF Fat Titz Hybrid(7): stop
<SF Fat Titz Hybrid(7): info depth 40 seldepth 51 multipv 1 score cp -432 nodes 192092073 nps 693150 hashfull 1000 tbhits 0 time 277129 pv f4e6 e5d7
<SF Fat Titz Hybrid(7): bestmove f4e6 ponder e5d7
>SF Fat Titz Hybrid(7): isready
<SF Fat Titz Hybrid(7): readyok

booot76 commented 1 year ago

Yes - looks like a bug. I will research it today.

booot76 commented 1 year ago

Yesterday i found the reason and fixed my structures. Thank you very much for your report again. Today engine plays test games. After it i will replace the binaries.

tissatussa commented 1 year ago

..Thank you very much for your report again..

it's my pleasure, and i'm glad you solved it !

btw. probably you'll not need this extra info, but i like to share it here : while solving another problem regarding Ponder (also a GitHub chess engine project) it seemed the "Ponder Hit" (!?) came "too quick" and so the application failed -- i'm not sure what this exactly means though .. HTH

booot76 / Booot-chess-engine

Ponder may contain a bug #22