Page 1 of 2

Stockfish_PA

Posted: Mon Jun 20, 2022 10:33 am
by Peter
Honoured Mr. Bernstein, dear Jeremy!

PA meaning persistent analysis was a SF- branch with learning- file, it's full name was SF_PA_GTB (gaviota tbs) using tbs at all as one of the first SF- versions doing so, and it was your code, so my question to you.

The experience- files coming up with some code from Kelly Kinyama since several years to me never worked as well as SF_PA's did compared to it's much lower strength way back then.

Could you imagine to try a revival of engine- learning- file for SF NNUE once in a while again?

Would be great for many users I guess, for my pov both of the old features, learning- file and full persistent hash (transposition tables?) like SF_PA then had, would be very fine tools of analysis, I on my own wouldn't care much about STC- game- playing for such purposes at all.

Thanks in advance for letting me know.

Best regards
Peter.

Re: Stockfish_PA

Posted: Mon Jun 20, 2022 2:05 pm
by Peter
BTW: editing Signature doesn't work at the moment and edit- time of posting is over
:mrgreen:

Re: Stockfish_PA

Posted: Mon Jun 20, 2022 6:51 pm
by deeds
Peter wrote:
Mon Jun 20, 2022 10:33 am
Honoured Mr. Bernstein, dear Jeremy!

PA meaning persistent analysis was a SF- branch with learning- file, it's full name was SF_PA_GTB (gaviota tbs) using tbs at all as one of the first SF- versions doing so, and it was your code, so my question to you.

The experience- files coming up with some code from Kelly Kinyama since several years to me never worked as well as SF_PA's did compared to it's much lower strength way back then.

Could you imagine to try a revival of engine- learning- file for SF NNUE once in a while again?

Would be great for many users I guess, for my pov both of the old features, learning- file and full persistent hash (transposition tables?) like SF_PA then had, would be very fine tools of analysis, I on my own wouldn't care much about STC- game- playing for such purposes at all.

Thanks in advance for letting me know.

Best regards
Peter.
Or you can try StockfishMZ (stockfish + learning), no doubt MZ can add a persistent hash feature on his engines.

Re: Stockfish_PA

Posted: Mon Jun 20, 2022 8:26 pm
by Peter
deeds wrote:
Mon Jun 20, 2022 6:51 pm
Or you can try StockfishMZ (stockfish + learning), no doubt MZ can add a persistent hash feature on his engines.
Thanks, he did already in his private engine Hypnos.
And there's this one feature in the engine with the disgusting name from Sopel, you'll probably know, which engine I mean
:)
Yet a learning file like the one of SF_PA's kind would be great to have too for me. MZ btw. made a revival of it way back then in Outskirts too, a SugaR_PA1.0 afair, but gave it up then again,

Re: Stockfish_PA

Posted: Wed Jun 29, 2022 11:55 am
by Indian15
ok I understand you are good in chatting but where is the engine with source code to download??

Re: Stockfish_PA

Posted: Thu Jun 30, 2022 3:13 pm
by Peter
Indian15 wrote:
Wed Jun 29, 2022 11:55 am
ok I understand you are good in chatting but where is the engine with source code to download??
Here's the latest thread I found about it, btw. not by forum- search, which doesn't seem to work for me here at the moment at all, I found the posting by Google

http://www.open-chess.org/viewtopic.php ... 067#p12799

SugaR_PA 1.0 is several years ago too, long before NNUE, pity I don't find it in Outskirts anymore neither and haven't anything stored on my actual PC's disks

Re: Stockfish_PA

Posted: Thu Jun 30, 2022 3:30 pm
by Peter
Just after edit time- out:

Google finds even more, finally yet some download- link for SF 4 PA_GTB here:

http://open-chess.org/viewtopic.php?p=1 ... 3ab#p19480

Direct download- link even worked still on my trial just now:

http://open-chess.org/download/file.php ... 3ee2e103ab

And a benchmark of the 64bit- compile at my 3.5GHz- Ryzen CPU:

Total time (ms) : 2611
Nodes searched : 8151340
Nodes/second : 3121922

Re: Stockfish_PA

Posted: Fri Jul 01, 2022 2:12 am
by sarona
Hi, Peter

Is not Stockfish TCEC6 PA_GTB the last build by Jeremy? This is the source I have been using.

see here:

Code: Select all

https://open-chess.org/viewtopic.php?f=7&t=2663
attachment still working:

Code: Select all

https://open-chess.org/download/file.php?id=597

Re: Stockfish_PA

Posted: Fri Jul 01, 2022 7:54 am
by Peter
sarona wrote:
Fri Jul 01, 2022 2:12 am
Is not Stockfish TCEC6 PA_GTB the last build by Jeremy? This is the source I have been using.
Thanks for that!
You're probably right, problem is, forum- search still doesn't work for me.
Bench of TCEC6 PA:

Total time (ms) : 2332
Nodes searched : 7307017
Nodes/second : 3133369

BTW, about MZ, there wasn't any SugaR_PA as I wrote wrongly, the name was Stockfish_PA_MZ 1.0,

Re: Stockfish_PA

Posted: Fri Jul 01, 2022 10:24 pm
by sarona
It looks like he joined the Komodo team after the release of Stockfish 7.

Unfortunately, I do not have Stockfish_PA_MZ 1.0. This must have been released during the MZ Chess Forum days.