New engine: Lumac0

[Dave Benson]

Hi,
I've been working on a chess engine, Lumac0, since January 2022.
It is open source and avalaible on Lichess, where it is rated around 1850-1900 points; it's UCI-compatible with no integrated book nor tablebases.

I'd like to see how well it can do on the CCRL lists, even though it's still quite weak... (I estimate it to be around 1500 points)

New releases will be posted on github, at least the most important ones.
I've also published the windows executable in the releases section, but the code is very simple to compile manually with g++.

I think the UCI interface works quite well, but any suggestion is higly appreciated .
Thanks for your attention, and please let me know if is there any other info that you may be interested in.

[Gabor Szots]

Hi Dave or Davide,

Which is your real name and where are you from?

[Gabor Szots]

I have just noticed that it uses at least 2 threads. That is unacceptable for me because I only test single CPU engines. Why can't it be set to 1?

[Dave Benson]

My real name is Davide and I'm Italian.

The engine uses 2 threads because it uses one thread for input/time checking and one for searching; if that's a problem I'll fix it as soon as possible.

[Gabor Szots]

My real name is Davide and I'm Italian.

The engine uses 2 threads because it uses one thread for input/time checking and one for searching; if that's a problem I'll fix it as soon as possible.

I have started a tournament with it because I can see that its CPU usage is mostly limited to 1 core, I guess the input polling and time checking is done only at certain intervals so maybe not a problem.

BTW, I see a 700 MB memory usage when there is 256 MB hash setting, why is that?

[Dave Benson]

Thank you for testing it; I have no idea why it is using so much RAM.
I'm sorry for all these inconveniences, as soon as I understand where the problem is I'll post the fixed version.

[Dave Benson]

There indeed was a problem in the computation of how much memory to allocate, I've fixed that and put the new version on GitHub as v12.15.

[Gabor Szots]

There indeed was a problem in the computation of how much memory to allocate, I've fixed that and put the new version on GitHub as v12.15.

OK, I'll replace the exes and rename 12.14 to 12.15 if otherwise there is no difference between them. Please confirm.

[Dave Benson]

Yes, the only difference between the two versions is the hash size handling bugfix.

[Gabor Szots]

Yes, the only difference between the two versions is the hash size handling bugfix.

Thanks. Now Arena displays 196 MB so I deduce hash size is 3/4 * 2exp(n) and is rounded downwards if the given value is greater.