Page 1 of 2

Monolith 1.0

Posted: Thu Oct 18, 2018 1:22 pm
by cimarronOST
I'm very proud to finally release Monolith 1.0 after almost 2 years of delights and frustration.
The engine is certainly nothing ground breaking but simply a plain old alpha-beta-searcher like hundreds of its kind,
but writing it has enriched those past 2 years immensely. To every non-programmer out there:
If you want to broaden your horizon and learn a programming language, there's no better way to get started than to
write a chess engine. Yes, it demands a great lot of time, but I regret no single hour of it.
Source and binaries can be found here: https://github.com/cimarronOST/Monolith/releases

Main changes to version 0.4:
-added support for multi-processors with simplified-ABDADA
-added support for syzygy endgame-tablebases
-added support for PEXT instruction
-added new search enhancements (SEE pruning, singular extension, LMR, aspiration window)
-added new evaluation terms
-tuned the evaluation automatically with Texel's tuning method
-simplified the code
-fixed a bug that caused crashes while moving backwards through the game

Strength estimate (bullet games):
Score of Monolith 1.0 vs Monolith 0.4: 6481 - 1470 - 2049 [0.751] 10000
Elo difference: 191.36 +/- 6.70

I'll also add Android compiles soon. A pity I'm always a couple of days too late with my releases for Graham's ongoing Division tourneys,
but I'm looking forward to the next series.

Have fun, that's what it's all about.

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 2:55 pm
by CMCanavessi
Wow, 200 elo over 0.4, what an improvement!!

BTW, the "pext" binary just hangs after depth 7 on my Ryzen processor. I know it's just very slow and that I should use popcnt one, but still, it shouldn't hang. Maybe other people with Ryzens can confirm the behaviour...

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 2:58 pm
by Gabor Szots
Splendid news, Jonas. Very nice improvement, and I congratulate you on your grit and determination. Also, it is good to hear that you have found immense reward in what you have achieved. Keep up the good work.

PS. Test gauntlet in progress...

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 3:02 pm
by Gabor Szots
CMCanavessi wrote: Thu Oct 18, 2018 2:55 pm Wow, 200 elo over 0.4, what an improvement!!

BTW, the "pext" binary just hangs after depth 7 on my Ryzen processor. I know it's just very slow and that I should use popcnt one, but still, it shouldn't hang. Maybe other people with Ryzens can confirm the behaviour...
Same here with an i5-4690k, so I am currently running the popcnt version.

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 3:34 pm
by Gabor Szots
Gabor Szots wrote: Thu Oct 18, 2018 2:58 pm Splendid news, Jonas. Very nice improvement, and I congratulate you on your grit and determination. Also, it is good to hear that you have found immense reward in what you have achieved. Keep up the good work.

PS. Test gauntlet in progress...

EDIT: That PS was a lie. None of my GUI's (Arena, F15, Sh13) could make Monolith move. I get warnings that tablebase/pawn limit is too low.

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 3:56 pm
by Gabor Szots
Gabor Szots wrote: Thu Oct 18, 2018 3:34 pm
Gabor Szots wrote: Thu Oct 18, 2018 2:58 pm Splendid news, Jonas. Very nice improvement, and I congratulate you on your grit and determination. Also, it is good to hear that you have found immense reward in what you have achieved. Keep up the good work.

PS. Test gauntlet in progress...

EDIT: That PS was a lie. None of my GUI's (Arena, F15, Sh13) could make Monolith move. I get warnings that tablebase/pawn limit is too low.
No luck even with the default parameters. As soon as book has been left, it stops moving. I am in the process of trying to determine what is wrong but time is running out on me soon.

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 4:01 pm
by Gabor Szots
Gabor Szots wrote: Thu Oct 18, 2018 3:56 pm
Gabor Szots wrote: Thu Oct 18, 2018 3:34 pm
Gabor Szots wrote: Thu Oct 18, 2018 2:58 pm PS. Test gauntlet in progress...

EDIT: That PS was a lie. None of my GUI's (Arena, F15, Sh13) could make Monolith move. I get warnings that tablebase/pawn limit is too low.
No luck even with the default parameters. As soon as book has been left, it stops moving. I am in the process of trying to determine what is wrong but time is running out on me soon.
EDIT: It seems the popcnt version is able to move if using default parameters.

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 6:49 pm
by cimarronOST
Gabor Szots wrote: Thu Oct 18, 2018 4:01 pm
Gabor Szots wrote: Thu Oct 18, 2018 3:56 pm
Gabor Szots wrote: Thu Oct 18, 2018 3:34 pm
Gabor Szots wrote: Thu Oct 18, 2018 2:58 pm PS. Test gauntlet in progress...

EDIT: That PS was a lie. None of my GUI's (Arena, F15, Sh13) could make Monolith move. I get warnings that tablebase/pawn limit is too low.
No luck even with the default parameters. As soon as book has been left, it stops moving. I am in the process of trying to determine what is wrong but time is running out on me soon.
EDIT: It seems the popcnt version is able to move if using default parameters.
Thanks for reporting this, I'll look into it asap. I have to admit that I didn't test the pext-version at all because I don't have a pext-compatible computer, but there seems to be something wrong also with the syzygy probing code. Sorry about that, I'm trying to reproduce and fix the problems.

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 8:06 pm
by tpoppins
No issues running the POPCNT compile under Cute Chess GUI so far. The new version easily beat v0.4 in the single 40/4 test game I ran, outsearching it by 3-4 plies in the middlegame. Configured to access 5-men Syzygy, it duly produced tbhits when expected.

Re: Monolith 1.0

Posted: Thu Oct 18, 2018 9:45 pm
by Graham Banks
Wow! Nice strength gain, Jonas. :)

This version might be way too strong for my current tourney. Will take a look.