Cheng 4.40

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

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
mar
Posts: 2298
Joined: Fri Nov 26, 2010 1:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Cheng 4.40

Post by mar » Sat Jan 02, 2021 8:17 pm

Hi all,

because I've made no progress since 4.40dev and because CEGT is already testing Guenther's compiles, I decided to make this an official release.
I've sent an e-mail to Emil so hopefully http://vlasak.biz/cheng will be updated soon.
for those who can't wait I also made a release on github: https://github.com/kmar/cheng4/releases/tag/4.40, just click the zip file.
special thanks to David Safranek (http://sx.rosada.cz) who runs a Cheng bot on Lichess, which motivated me to work a bit on Cheng once more.

happy New year and have fun

Martin
Martin Sedlak

Gabor Szots
Posts: 752
Joined: Sat Jul 21, 2018 5:43 am
Location: Szentendre, Hungary
Full name: Gabor Szots

Re: Cheng 4.40

Post by Gabor Szots » Sat Jan 02, 2021 9:00 pm

mar wrote:
Sat Jan 02, 2021 8:17 pm
Hi all,

because I've made no progress since 4.40dev and because CEGT is already testing Guenther's compiles, I decided to make this an official release.
I've sent an e-mail to Emil so hopefully http://vlasak.biz/cheng will be updated soon.
for those who can't wait I also made a release on github: https://github.com/kmar/cheng4/releases/tag/4.40, just click the zip file.
special thanks to David Safranek (http://sx.rosada.cz) who runs a Cheng bot on Lichess, which motivated me to work a bit on Cheng once more.

happy New year and have fun

Martin
Thank you Martin. Nice surprise. I hope your motivation will last very long.
Gabor Szots
CCRL testing group

Angle
Posts: 212
Joined: Sat Oct 31, 2020 12:04 am
Full name: Aleksey Glebov

Re: Cheng 4.40

Post by Angle » Mon Jan 04, 2021 7:36 pm

mar wrote:
Sat Jan 02, 2021 8:17 pm
because I've made no progress since 4.40dev and because CEGT is already testing Guenther's compiles, I decided to make this an official release.
I've sent an e-mail to Emil so hopefully http://vlasak.biz/cheng will be updated soon.
for those who can't wait I also made a release on github: https://github.com/kmar/cheng4/releases/tag/4.40, just click the zip file.
special thanks to David Safranek (http://sx.rosada.cz) who runs a Cheng bot on Lichess, which motivated me to work a bit on Cheng once more.
Hi, Martin! Any changes compared to 4.40 dev?
Incredibly fast systems miscount incredibly fast.

mar
Posts: 2298
Joined: Fri Nov 26, 2010 1:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Re: Cheng 4.40

Post by mar » Mon Jan 04, 2021 7:49 pm

Angle wrote:
Mon Jan 04, 2021 7:36 pm
Hi, Martin! Any changes compared to 4.40 dev?
No changes, bench is the same. So 4.40 dev = 4.40
Martin Sedlak

Frank Quisinsky
Posts: 5467
Joined: Wed Nov 18, 2009 6:16 pm
Location: Trier, Germany
Full name: Frank Quisinsky
Contact:

Re: Cheng 4.40

Post by Frank Quisinsky » Mon Jan 04, 2021 10:16 pm

Hi Martin,

do you saw the "short won game-stats" from FCP Qualify Tourney-2021?
That's what I mean why Cheng is all the time a popular program, we discuss about it in "TalkChess-Tourney selection".

50-100 Elo more and you engine in in TOP-40.

The same problem Spark by Allard Siemelink have in the past.
A fantastic playing style in mid-games, same Cheng have, but in endgames the strength goes back.

All in all ...
Cheng is one of my favorite programs and maybe you can create a version for DGT-Pi?
Chess computer lovers will like Cheng a lot and of course I am a more happy man with Cheng on DGT-Pi for self playing.

Best for you and congratulation for the great improvements!
Frank
I like computer chess!

mar
Posts: 2298
Joined: Fri Nov 26, 2010 1:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Re: Cheng 4.40

Post by mar » Mon Jan 04, 2021 11:33 pm

what type of endgames do you mean? pawnless? this is what I saw as well - trying to win drawish endgames.
there're more problems than this, of course

improving 50-100 elo is not easy, I've spent several months on this new version, played literally hundreds of thousands of bullet games.
the result? I guess 30 elo at best, but we'll see...

as for DGT Pi - I don't have a DGT board so no idea, but since Cheng is open source, I guess someone else might try to compile it for you?
Martin Sedlak

Frank Quisinsky
Posts: 5467
Joined: Wed Nov 18, 2009 6:16 pm
Location: Trier, Germany
Full name: Frank Quisinsky
Contact:

Re: Cheng 4.40

Post by Frank Quisinsky » Tue Jan 05, 2021 12:10 am

Hi Martin,

style is more important as Elo, for myself all is OK with Cheng!
So the work you do on Cheng 4.40 must be good!

I saw a lot of Cheng games during the FCP Qualify Tournament is still running.
Should be clear, I like your engine and watching!

Not easy to explain:
A delay process with pawn endgames.
Or maybe in better english ... a sneaky / creeping process with pawn-structurs in endgames.

Often I am thinking after Cheng switched rooks, bishops or knights the pawn-positions after ... must be lose.
Often I am thinking why Cheng don't try to bring rooks on better positions.
2-line, 7-line on the board for rooks, the rooks are not aggressively enough in endgames.

But in reality I am thinking that most of the others with the same playing strength, or a bit stronger as Cheng, can build more plys in endgames and they are faster and can see more.

Again, the mid-game is very nice and the reason that I like your engine a lot!

I am also not a raspberry expert.
The admin from chess computer info send me after Wasp Release a new image!
Yesterday in the evening I got it, version 3 from Picochess is fantastic!!
More as fantastic only ... a sensation!

All the older chess computers can run and with the 22 levels Wasp build I can play against Wasp on DGT board with 1500-2696 Elo in 22 levels.

Your Cheng will be great for such things (chess computer community will like it, not I alone) because the style Cheng produced is human-like (strong in mid-games and lost strength in endgames). So humans will have a chance to win without artfull actions ... like to try to make a strong chess program weaker. The other way DGT go with the chess computers DGT offer, very bad to use Stockfish here!

Best
Frank

To build now an Cheng endgame monster for more Elo ...
I am quite sure, we don't need engine 100 strong in endgames!!!
Much more interesting is the way you go, forget the endgame and try to make the mid-game stronger and stronger! The own face Cheng have is fantastic and most important is to hold it. The reaon why I like the work John Stanback do since some years with all the energy day by day John give Wasp. And your Cheng is also ... especially!
Last edited by Frank Quisinsky on Tue Jan 05, 2021 12:25 am, edited 1 time in total.
I like computer chess!

petero2
Posts: 612
Joined: Mon Apr 19, 2010 5:07 pm
Location: Sweden
Contact:

Re: Cheng 4.40

Post by petero2 » Tue Jan 05, 2021 12:23 am

mar wrote:
Mon Jan 04, 2021 11:33 pm
as for DGT Pi - I don't have a DGT board so no idea, but since Cheng is open source, I guess someone else might try to compile it for you?
The following link may be interesting for people trying to port multithreaded C/C++ code to ARM CPUs:

https://docs.microsoft.com/en-us/cpp/cp ... w=msvc-160

User avatar
CMCanavessi
Posts: 929
Joined: Thu Dec 28, 2017 3:06 pm
Location: Argentina

Re: Cheng 4.40

Post by CMCanavessi » Tue Jan 05, 2021 1:06 am

Any estimates of improvement against 4.39 ?
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls

mar
Posts: 2298
Joined: Fri Nov 26, 2010 1:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Re: Cheng 4.40

Post by mar » Tue Jan 05, 2021 1:16 am

petero2 wrote:
Tue Jan 05, 2021 12:23 am
The following link may be interesting for people trying to port multithreaded C/C++ code to ARM CPUs:

https://docs.microsoft.com/en-us/cpp/cp ... w=msvc-160
thanks Peter.
I understand that ARM has a weaker memory model than x86, but since Cheng ran on Adroid ARMs, I guess it might be no real issue.
I use volatile in a couple of places where I should be using atomics (like search abort flag), search commands are using pthread mutex, so this is no problem; still I allow races on TT, but since it won't stay in caches for long, that should be probably fine as well.
also, I'm not sure if there were any changes in the memory model in AARCH64, so no idea (previous Android ARM compiles were 32-bit only).

I didn't include an Android build this time, because of what Aart wrote about the new security changes in Android - assuming it only affects running engines from SSDs (for now), I probably should've added an Android compile - but since I don't own an Android device at the moment (plus I was lazy to set up NDK), I decided to pass on Adroid this time. Someone else can build Android binaries for sure.

what I meant is I don't know what OS DGT Pi runs - some flavor of Linux/Android I guess? so I've no idea if an Android binary would actually do for DGT Pi?
Martin Sedlak

Post Reply