Page 4 of 6

Re: It's here ...Honey X6

Posted: Thu Oct 10, 2019 6:24 am
by carldaman
Dann Corbit wrote:
Thu Oct 10, 2019 12:36 am

I called it polygon {many-sided} because it uses all of the features except it does not have the blue stuff yet.

Updated file at the same location, last version did not log the EPD position:
https://drive.google.com/open?id=1tLQiV ... ICHiAhzJZ0
Thanks for the compile, Dan. What do you mean by 'not having the blue stuff yet'?

Re: It's here ...Honey X6

Posted: Thu Oct 10, 2019 6:36 am
by MikeB
carldaman wrote:
Thu Oct 10, 2019 6:24 am
Dann Corbit wrote:
Thu Oct 10, 2019 12:36 am

I called it polygon {many-sided} because it uses all of the features except it does not have the blue stuff yet.

Updated file at the same location, last version did not log the EPD position:
https://drive.google.com/open?id=1tLQiV ... ICHiAhzJZ0
Thanks for the compile, Dan. What do you mean by 'not having the blue stuff yet'?
There was a conflict with using 'ifdef Blue' with windows.h. On misc.cpp, doing something like this near the top will fix it:

Code: Select all

#ifdef Blue
#undef Blue
#define BlueAgain
#endif
#include <windows.h>
#ifdef BlueAgain
#define Blue
#endif
This only impacted the "bluefish" exe's.

Re: It's here ...Honey X6

Posted: Thu Oct 10, 2019 2:27 pm
by Spliffjiffer
MikeB wrote:
Thu Oct 10, 2019 6:01 am
MikeB wrote:
Tue Oct 08, 2019 7:42 pm
well if you have a macOS computer anyway , or if you can compile your own

other OS flavors will be released upon receipt.

all versions have all bell and whistles, including play by FIDE Elo (or CCRL Elo) and improved adaptive and variety play
FD versions have 'fortress detection' code written Joe Ellis.
Bluefish and Blue-Honey versions have Bluefish evaluation code written by Eelco de Groot.

https://github.com/MichaelB7/Stockfish/releases/tag/X6

that's right - 8 stockfish derivative flavors , including

current-dev-Stockfish
current-dev-Stockfish-FD
Honey X6
Honey-FD
Blue-Honey
Blue-Honey-FD
Bluefish
BlueFish-FD

there's build script in the src folder, on my machine it takes 62 seconds to build all 8 flavors, just over two minutes for PGO flavors ( all macOS versions were PGO optimized)

why "X" ? , it signifies that this version is related to Stockfish 10 and development code added since the release of Stockfish 10.

have fun!
come and get it ... all flavors now available, scroll to the bottom for the binaries...

https://github.com/MichaelB7/Stockfish/releases/tag/X6

macSO users should re-download ( also bmi2 was added)

Also, please be sure to thank Alan Cooper and Lucas Monge for all their work in making the exe's available for you!
What a helpful Honey-beast, especially for long-time-analysis...im amazed and happy :-)
i want to say thank you to:
Dann Corbit for his "poly-version"...and
Joe Ellis for his provided FD-code (a dream came true as an analyzer)...and
Eelco de Groot for his bluefish-code made available (reminds me of the old "aggressivness/coverdice"-days :-) )...and
Alan Cooper and Lucas Monge for the compiles they always provide...and
the SF-team, everybody knows why...and
of course MikeB as the creator, always kind person and author !

thanks so much, you are great :-)
beyond that i wish joy and peace for everybody

peer

Re: It's here ...Honey X6

Posted: Fri Oct 11, 2019 2:23 am
by MikeB
Spliffjiffer wrote:
Thu Oct 10, 2019 2:27 pm
MikeB wrote:
Thu Oct 10, 2019 6:01 am
MikeB wrote:
Tue Oct 08, 2019 7:42 pm
well if you have a macOS computer anyway , or if you can compile your own

other OS flavors will be released upon receipt.

all versions have all bell and whistles, including play by FIDE Elo (or CCRL Elo) and improved adaptive and variety play
FD versions have 'fortress detection' code written Joe Ellis.
Bluefish and Blue-Honey versions have Bluefish evaluation code written by Eelco de Groot.

https://github.com/MichaelB7/Stockfish/releases/tag/X6

that's right - 8 stockfish derivative flavors , including

current-dev-Stockfish
current-dev-Stockfish-FD
Honey X6
Honey-FD
Blue-Honey
Blue-Honey-FD
Bluefish
BlueFish-FD

there's build script in the src folder, on my machine it takes 62 seconds to build all 8 flavors, just over two minutes for PGO flavors ( all macOS versions were PGO optimized)

why "X" ? , it signifies that this version is related to Stockfish 10 and development code added since the release of Stockfish 10.

have fun!
come and get it ... all flavors now available, scroll to the bottom for the binaries...

https://github.com/MichaelB7/Stockfish/releases/tag/X6

macSO users should re-download ( also bmi2 was added)

Also, please be sure to thank Alan Cooper and Lucas Monge for all their work in making the exe's available for you!
What a helpful Honey-beast, especially for long-time-analysis...im amazed and happy :-)
i want to say thank you to:
Dann Corbit for his "poly-version"...and
Joe Ellis for his provided FD-code (a dream came true as an analyzer)...and
Eelco de Groot for his bluefish-code made available (reminds me of the old "aggressivness/coverdice"-days :-) )...and
Alan Cooper and Lucas Monge for the compiles they always provide...and
the SF-team, everybody knows why...and
of course MikeB as the creator, always kind person and author !

thanks so much, you are great :-)
beyond that i wish joy and peace for everybody

peer
Definitely a team effort - and none of this happens unless Stockfish was set up the way it is - Tord, Marco and the scores of others who are either providing patches, providing hardware, the maintainers and all the users who report bugs - it really is a thing of beauty, how this has worked out the last ten years and my hope is that it is still here way after I'm gone.

Re: It's here ...Honey X6

Posted: Fri Oct 11, 2019 1:35 pm
by maxdeg
MikeB wrote:
Fri Oct 11, 2019 2:23 am
way after I'm gone.
Not yet Mike :D - I need to know what settings you suggest for LTC analysis using any variety of Honey .. please .. :)
An image of the parameter file would be useful.

+3 for the Honey macOS compiles.

Re: It's here ...Honey X6

Posted: Fri Oct 11, 2019 2:14 pm
by Paloma
by MikeB
This only impacted the "bluefish" exe's.
What exactly does this mean.

Can I use this "blue" exe's only limitedly?

Is this a bug? Of course not. But what else is this exactly?

Re: It's here ...Honey X6

Posted: Fri Oct 11, 2019 6:37 pm
by Dann Corbit
I updated my polygon build to include the blue stuff.
I changed Blue to Bule everywhere in the program except for text strings.
It's a dyslexic thing. I do that sort of thing by accident sometimes, so it seemed funny to me.

https://drive.google.com/open?id=1tLQiV ... ICHiAhzJZ0

I ran it on some really tough positions last night, and the outcome looks extremely promising.
If you use my version, don't forget to change my strange defaults like exit after bench and logging to a file.

Re: It's here ...Honey X6

Posted: Fri Oct 11, 2019 6:39 pm
by Dann Corbit
Paloma wrote:
Fri Oct 11, 2019 2:14 pm
by MikeB
This only impacted the "bluefish" exe's.
What exactly does this mean.

Can I use this "blue" exe's only limitedly?

Is this a bug? Of course not. But what else is this exactly?
To answer your question, on Windows systems, Blue is a macro in the header file Windows.h, so all sorts of sparks would fly when you tried to do Blue builds.

As for what does Blue mean, that means that it has some of Eelco's evaluation parameters.

Re: It's here ...Honey X6

Posted: Fri Oct 11, 2019 6:50 pm
by Dann Corbit
There is one significant evaluation change that I make for null move reductions.
I have a logistic S-curve function to dampen pruning when mobility is cramped and to increase pruning when mobility is enormous.
IOW, if you only have 3 possible moves, cut pruning way back. And if you have 140 possible moves, prune like mad. In the function below, maxmove is the count of possible moves from this position.

double dR = ((835.0 + 70.0 * ddepth ) * c256inv + std::min((double(eval - beta) * pvinv), 3.0)) * logistic_mobility_score(maxmove);

Re: It's here ...Honey X6

Posted: Sat Oct 12, 2019 11:53 pm
by carldaman
Thanks +1 for the new release. :D The style is there as before, but contempt seems to be broken. I use centipawn score, and the scores don't reflect any sort of activated contempt, no matter what it's set to.