Page 4 of 5

Re: Cool Iris

Posted: Sat Dec 16, 2023 11:41 am
by Solista
Cool Iris 11.80 with external network (avx2, avx512, bmi2, sse41, modern, source code):

https://pixeldrain.com/u/RbFpk7m9

Thank You Anton.

Re: Cool Iris

Posted: Sun Dec 17, 2023 2:50 am
by Anton101
Thank you so much Eduard !

I forget to left the downloads, Im using right now your builts!

I wish you the best!

Best regards;

Re: Cool Iris

Posted: Sun Dec 17, 2023 11:10 am
by Solista
Hello Anton,
I happened to see this line in the Timeman of Cool Iris 11.80 (I only looked at the code because the SF developers removed the slow mover):

Why did you change the code? Is time management of Cool Iris 11.80 better?

Codes:

Cool Iris 11.80:
optScale = std::min(0.0120 + std::pow(ply + 3.1, 0.44) * optConstant,
0.21 * limits.time[us] / double(timeLeft))
* optExtra;
maxScale = std::min(6.9, maxConstant + ply / 12.2);

Cool Iris 11.70:
optScale = std::min(0.0120 + std::pow(ply + 3.3, 0.44) * optConstant,
0.2 * limits.time[us] / double(timeLeft))
* optExtra;
maxScale = std::min(6.8, maxConstant + ply / 12.2);

Re: Cool Iris

Posted: Sun Dec 17, 2023 4:36 pm
by Anton101
Hi Eduard!

Because Cool Iris is CorChess and the engine works better with VLTC, I decided to take advantage of the Fauzi patch, intended for this type of control.

Every week I receive emails where some users have mentioned that they use the engine in long TC's for some tests, for example 10+5, 15 +2... etc., so, I just took advantage of this patch

On the other hand I kept "slow mover" for server users I know they will have no problem with the engine in control times 1+0, 3+0 or 5+0,

By the way, the change was made in timeman.cpp and search.cpp

I leave you the link to the patch:
https://tests.stockfishchess.org/tests/ ... cf40aaf914

Best regards;

Re: Cool Iris

Posted: Sun Dec 17, 2023 7:36 pm
by Solista
Thanks Anton! Question:

This page of Stockfish testing
https://groups.google.com/g/fishcooking-results?pli=1

is no longer updated.

Is this the only site with new tests:
https://tests.stockfishchess.org/tests/ ... ess_only=1

Re: Cool Iris

Posted: Sun Dec 17, 2023 10:29 pm
by Anton101
This is the link that I review:

https://tests.stockfishchess.org/tests

You welcome.

Re: Cool Iris

Posted: Sun Jan 21, 2024 4:30 pm
by Solista
Hello Anton, there has been no new Corchess for 14 days. Corchess may be dead. The Stockfish developers changed the entire SF architecture on January 11th. The Corchess Code, which has existed in this form for years, is history. That's why I'm going my own way now. How about Cool Iris for you? Maybe developed I. Ivec a new corchess code?

Re: Cool Iris

Posted: Sun Jan 21, 2024 7:55 pm
by Anton101
Hi Eduard!

I have concentrated more effort on my personal engines (Dark SisTer and SpecTral), on the other hand the Elo gains are still minimal, there will be a new Cool Iris, but I will wait for SF 16.1 In the meantime I will continue doing private tests.
I prefer Dark SisTer and SpecTral do not handle experience files, Cool Iris will continue to use experience files and will support CTG/Bin books. I don't like that the experience file continues using SugaR exp Tool, but if there is no option, it will continue that way.

If you want a copy of DS and SpecTral, just give me an email and flavor, please (avx2, bmi2 or modern).

Best regards;

Re: Cool Iris

Posted: Mon Jan 22, 2024 6:12 pm
by Solista
Thanks Anton! I have now taken a closer look at the new SF source code. It is very easy to implement the corchess code. That's why I think I Ivec will release a new Corchess soon. Where did you get the information that Stockfish 16.1 is coming out soon?

Re: Cool Iris

Posted: Mon Jan 22, 2024 9:14 pm
by Anton101
There are rumors on discord that it is very likely that February will be the chosen month. Therefore, nothing happens if I am patient waiting for an important news. (in patches).

And by the way, I insist, I liked the modification in evaluate.cpp (Leptir) I don't know where these numbers came from, but they strong and effective.

Congratulaions, again.

Best regards;