Slightly smoother stockfishes

[Dann Corbit]

http://cap.connx.com/chess-engines/new- ... 32.exe.bz2
http://cap.connx.com/chess-engines/new- ... 64.exe.bz2

I did not test them at all yet.

[royb]

Dann,

Any idea when the source for these will be available? I need to compile for Linux...

Thanks,

Roy

[Dann Corbit]

Dann,

Any idea when the source for these will be available? I need to compile for Linux...

Thanks,

Roy

Collect the regular stockfish source and add this:
http://cap.connx.com/chess-engines/new-approach/sfc.zip
There are two modifed C++ files in the zip.
Do not forget to define a macro of:
SMOOTH_REDUCTION
(you can find how it is used in search.cpp at line 1394).

[royb]

Dann,

I'll have to wait then. I would not know how to define a macro if my life depended on it.

Roy

[zullil]

http://cap.connx.com/chess-engines/new- ... 32.exe.bz2
http://cap.connx.com/chess-engines/new- ... 64.exe.bz2

I did not test them at all yet.

While I understand the excitement of making a change that seems to significantly enhance the strength of a program, I wonder if this episode might have been handled differently. We now have a number of different Windows binaries of Stockfish-1.6*, and it's not clear (to me at least) what source is producing them. Perhaps these should only have been released in packages that contained the source code that generated the binaries. Although I can run the binaries on my Mac via wine, the source would be better.


In any case, thanks for experimenting and strengthening Stockfish.

[Dann Corbit]

Dann,

I'll have to wait then. I would not know how to define a macro if my life depended on it.

Roy

If you use GCC, you can supply it on the command line.

Or in the top of the file you can add this line:

#define SMOOTH_REDUCTION 1

[Dann Corbit]

http://cap.connx.com/chess-engines/new- ... 32.exe.bz2
http://cap.connx.com/chess-engines/new- ... 64.exe.bz2

I did not test them at all yet.

While I understand the excitement of making a change that seems to significantly enhance the strength of a program, I wonder if this episode might have been handled differently. We now have a number of different Windows binaries of Stockfish-1.6*, and it's not clear (to me at least) what source is producing them. Perhaps these should only have been released in packages that contained the source code that generated the binaries. Although I can run the binaries on my Mac via wine, the source would be better.


In any case, thanks for experimenting and strengthening Stockfish.

I only changed two files. Here they are:
http://cap.connx.com/chess-engines/new-approach/sfc.zip
You will have to define a macro called SMOOTH_REDUCTION to incorporate the changes.

[Shaun]

http://cap.connx.com/chess-engines/new- ... 32.exe.bz2
http://cap.connx.com/chess-engines/new- ... 64.exe.bz2

I did not test them at all yet.

Dann,

I assume this is a further code tweak from the title - please change the name - we have now got another 1.6s from different code???

Shaun

[royb]

Dann,

I'll have to wait then. I would not know how to define a macro if my life depended on it.

Roy

If you use GCC, you can supply it on the command line.

Or in the top of the file you can add this line:

#define SMOOTH_REDUCTION 1

I added the line above to search.cpp and then did a 'make' with this output:

<snip...>

g++ -O1 -msse -DNDEBUG -g -Wall -fno-exceptions -fno-rtti -c -o movegen.o movegen.cpp
g++ -O1 -msse -DNDEBUG -g -Wall -fno-exceptions -fno-rtti -c -o history.o history.cpp
g++ -O1 -msse -DNDEBUG -g -Wall -fno-exceptions -fno-rtti -c -o movepick.o movepick.cpp
g++ -O1 -msse -DNDEBUG -g -Wall -fno-exceptions -fno-rtti -c -o search.o search.cpp
search.cpp: In function ‘Value<unnamed>::search(Position&, SearchStack*, Value, Depth, int, bool, int, Move)’:
search.cpp error: ‘max’ was not declared in this scope
value.h: At global scope:
value.h:126: warning: ‘TempoValue’ defined but not used
history.h:70: warning: ‘HistoryMax’ defined but not used
make[2]: *** [search.o] Error 1
make[2]: Leaving directory `/home/roy/chess/test-new/stockfish_1.6'
make[1]: *** [gcc] Error 2
make[1]: Leaving directory `/home/roy/chess/test-new/stockfish_1.6'
make: *** [default] Error 2

Any pointers are greatly appreciated!

Roy

[Dann Corbit]

http://cap.connx.com/chess-engines/new- ... 32.exe.bz2
http://cap.connx.com/chess-engines/new- ... 64.exe.bz2

I did not test them at all yet.

Dann,

I assume this is a further code tweak from the title - please change the name - we have now got another 1.6s from different code???

Shaun

Yes, you are right, and it is bad orgainization by me.
The current idea is in a very primitive state and I guess that it will be rapidly improved anyway by systematic analysis. Once Dr. Hyatt's cluster awakens, the idea will have the ever loving stuffings pounded out of it. I am looking forward to that with great longing.

At any rate, all of my builds are totally unofficial since I am not a member of the stockfish team. I just thought it was time that I shared this idea (and I am not very good at explaining things so I just tossed in a patch of code).

The only way to differentiate my slightly smoother version is by the date (it hash today's date on it). I do not know if it is better or worse. I am testing it right now.

The source patch archive on my ftp site has the latest version source in it also.