Gödel 4.4.5 released

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

Moderators: hgm, Rebel, chrisw

Gabor Szots
Posts: 1362
Joined: Sat Jul 21, 2018 7:43 am
Location: Szentendre, Hungary
Full name: Gabor Szots

Re: Gödel 4.4.5 released

Post by Gabor Szots »

jvazgo wrote: Thu Sep 27, 2018 12:06 am
Gabor Szots wrote: Tue Sep 25, 2018 12:24 pm Thanks Juanma.

May I ask you how to set hash size in WB mode? Whatver value I set in the .rc file Arena always reports 91 MB.
In Arena just press CTRL + 1 and set the hash size.
You must start a new game to see the change of the memory size in the status bar.
Remember that you can't play against Gödel in Winboard by the moment and that the book isn't working.

Thanks Gabor.
Yes, that works when Gödel is set to UCI but does not help when it is set to WB.
Gabor Szots
CCRL testing group
tpoppins
Posts: 919
Joined: Tue Nov 24, 2015 9:11 pm
Location: upstate

Re: Gödel 4.4.5 released

Post by tpoppins »

Changing options on the config screen does work in Winboard mode, you just need to start the engine analyzing or playing for it to take effect.
The hash size Arena (v3.5.1 at least) displays is nearly random: when it shows 3 it could actually be 4096; and when the hash is 128 it might show 12 -- always check with Task Manager, Process Explorer or the like instead.

What I find interesting is that, unlike Arena, Cute Chess GUI only shows the useless "cores" (Godel is non-SMP, isn't it?) option on Godel's config screen in Winboard mode. None of the options present when in UCI mode - "Hash", "Pawn hash", "Post ply" and "OwnBook" -- show up. Of course, why anyone would want to run Godel in Winboard mode under CC is a moot point; nevertheless it's interesting from the technical point of view.

My guess would be that Godel sends out the batch of "feature option" commands too early, right after "xboard". Perhaps if it waited for "protover 2" first then these options would show up.

As for the features it does send in response to the protover command, I'd like to point out two oddities there. The first "smp=1" is wrong if the engine is not SMP. That's what creates that misleading "cores" option that shouldn't be there in the GUI. The second one is "memory=0". These two commands and their values apparently got swapped at some point, because "smp" is enabled but nonfunctional and "memory" works but is disabled. If you change it to "memory=1" the corresponding option shows up on the engine's config screen, and that option really works, according to Task Manager.

Hope this info will be of some assistance.
Tirsa Poppins
CCRL
User avatar
Guenther
Posts: 4605
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: Gödel 4.4.5 released

Post by Guenther »

tpoppins wrote: Thu Sep 27, 2018 10:46 pm
My guess would be that Godel sends out the batch of "feature option" commands too early, right after "xboard". Perhaps if it waited for "protover 2" first then these options would show up.
Is it possible that 'feature options' for xboard are just not implemented in cutechess yet?
The cores command which is shown in cute was 'invented' earlier than feature options by HGM.
(But still may be in 2016 or so)

I can confirm that only cores show up as xboard option in cutechess,
but in WB all options show up!
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica + AlexChess + Eduard + Sylwy
tpoppins
Posts: 919
Joined: Tue Nov 24, 2015 9:11 pm
Location: upstate

Re: Gödel 4.4.5 released

Post by tpoppins »

Of course they are, CC wasn't written last week, you know. ;)

Here is its config screen for Scorpio 2.8:

Image

and here is Scorpio's console output for comparison:

Code: Select all

feature done=0
ht 33554432 X 16 = 512.0 MB
eht 1048576 X 8 = 8.0 MB
pht 32768 X 24 = 0.8 MB
treeht 3355392 X 40 = 128.0 MB
processors [1]
xboard
protover 2
feature name=1 myname="Scorpio_2.8"
feature sigint=0 sigterm=0
feature setboard=1 draw=0 colors=0
feature smp=0 memory=0
feature option="log -check 0"
feature option="clear_hash -button"
feature option="resign -spin 1000 100 30000"
feature option="cores -spin 1 1 64"
feature option="ht -spin 512 1 131072"
feature option="eht -spin 8 1 16384"
feature option="pht -spin 1 1 256"
feature option="egbb_path -path /home/dabdi/egbb/"
feature option="egbb_cache_size -spin 32 1 16384"
feature option="egbb_load_type -spin 3 0 3"
feature option="egbb_depth_limit -spin 6 0 64"
feature option="egbb_ply_limit_percent -spin 75 0 100"
feature option="smp_type -combo *YBW /// ABDADA /// SHT "
feature option="smp_depth -spin 8 1 10"
feature option="use_singular -check 1"
feature option="use_probcut -check 0"
feature option="singular_margin -spin 32 0 1000"
feature option="probcut_margin -spin 195 0 1000"
feature option="aspiration_window -spin 10 0 100"
feature option="UCTKmin -spin 30 0 100"
feature option="UCTKmax -spin 30 0 100"
feature option="reuse_tree -check 1"
feature option="backup_type -combo *MINMAX AVERAGE"
feature option="frac_alphabeta -spin 100 0 100"
feature option="mcts_strategy_depth -spin 15 0 100"
feature option="montecarlo -check 0"
feature option="treeht -spin 127 0 131072"
feature done=1
Tirsa Poppins
CCRL
jvazgo
Posts: 62
Joined: Tue Nov 22, 2011 9:18 am
Location: Sevilla, Spain.
Full name: Juan Manuel Vázquez

Re: Gödel 4.4.5 released

Post by jvazgo »

Hi Tirsa,

You were right, I was sending the "feature option" commands too early before receiving the protover 2 command.
The bug is solved sending all the "feature options" before the "feature done=1" command as you do in your example.
I didn't follow the winboard protocol version 2 correctly.

Thank you very much for your help, your Scorpio's output example was very useful.
Juanma.
Gabor Szots
Posts: 1362
Joined: Sat Jul 21, 2018 7:43 am
Location: Szentendre, Hungary
Full name: Gabor Szots

Re: Gödel 4.4.5 released

Post by Gabor Szots »

tpoppins wrote: Thu Sep 27, 2018 10:46 pm Of course, why anyone would want to run Godel in Winboard mode under CC is a moot point
Tirsa, when an engine is both WB and UCI and the GUI is WB and UCI capable as well, why do you think that UCI mode is the preferable selection?
Gabor Szots
CCRL testing group