Stockfish 3 PA_GTB

[Vael Jean-Paul]

Welcome to Git (version 1.8.1.2-preview20130201)
...

jpqy@JPQY-SAMSUNG /c/users/jpqy/Stockfish/src/kyotocabinet (master)
$ ./configure
#================================================================
# Configuring Kyoto Cabinet version 1.2.76.
#================================================================

...

checking pthread.h usability... yes
checking pthread.h presence... yes
checking for pthread.h... yes
checking regex.h usability... no
checking regex.h presence... no
checking for regex.h... no
configure: error: regex.h is required
jpqy@JPQY-SAMSUNG /c/users/jpqy/Stockfish/src/kyotocabinet (master)
$

I make the list short..and i get a new error?

i was using the 4.7.2 now updated to 4.7.4 and pthread is accepted..

[Jeremy Bernstein]

First check out the MINGW_COMPILE.txt file in the kyotocabinet directory, where I documented what I had to do in order to get kyotocabinet to build due to a missing regex library.

[Vael Jean-Paul]

That is Chinees for me..i need step by step help here..

I still use Git Bash to do this steps..right!
i have download these mingw-libgnurx-2.5.1 and copy them to that folder where the ones before are located.
What do now..i go with Git Bash to my MinGW directory first 64bit then 32bit and do this steps under..right!

If this works ,i come back for next step..

Compiling Kyoto Cabinet with MinGW

Compile and install regex (in src/kyotocabinet/mingw-libgnurx-2.5.1, or from http://sourceforge.net/projects/mingw/f ... gex-2.5.1/). This is a little tricky because you'll need versions for 32- and 64-bit MinGW. So first, using 64-bit MinGW:

> make clean
> ./configure --prefix=/mingw/mingw64/x86_64-w64-mingw32 [your path may be different]
> make
> make install

Then using 32-bit MinGW:

> make clean
> ./configure --prefix=/mingw/mingw32/i686-w64-mingw32
> make
> make install

Now Kyoto Cabinet. Stockfish doesn't use the compression features of Kyoto Cabinet, so we're not compiling zlib into it. If you want, you can build and install zlib as above and remove the --disable-zlib flag below.

32-bit:
> ./configure --disable-zlib --enable-static
> make

64-bit: this is slightly more complicated because of type-casting problems
> ./configure --disable-zlib --enable-static
now edit the makefile and add the -fpermissive flag to CXXFLAGS
> make

Congratulations, you did it.

[Vael Jean-Paul]

It doesn't work for me..

Git Bash do nothing when i type: make clean
cpu runs at 25% just stay there till i close Git Bash
The question is also in which directory to start..i tried on different places but he do nothing..

So,i'am lost here

Thanks for your help..this is really for people who can programming or understand what to do..
I need a step by step quide..
For sure there is a lot work to do to make these programs more use friendly..when you know it ,it' always easy..

I replace everything so that i can still compile the normal Stockfish sources..

Thanks for your time..meaning was that if i can do this ,then i can choose any Stockfish source to integrade these GTB bases myself..

Kind regards,
JP.

[User923005]

It seems that there is some problems for people to build the file.
Here is the version that I built using visual studio.
I had to change some source code a little bit (e.g. the CPU lookup) and a few other details.
It seems to work well for me.

[Jeremy Bernstein]

If you can compile using VS.NET with Dann's version, that's great. Mingw compiles are somewhat faster, though. That said, I really don't have the time right now to walk you through the process (sorry to be blunt, but I'm really busy with other stuff right now) -- the documentation I provided for kyotocabinet should be sufficient to get you building, provided you understand how the mingw toolchain works.

Good luck
Jeremy

[BorgKing]

I had to run chkdisk this morning again, after having SF 008 play during the night in Arena. No blue screen event yet though

To the programmers here, does any of the following make sense to you regarding what the cause may be? Is it in SF 008 or Arena or in a combination of them?

damaged attribute post (48, "") - [this one seems to be recurring every time this happens]

damaged file post segment 150176

name giving error 121764

restoring green2.png and GREEN24.png


There were some more, but these were the only ones I managed to get written down...

[Jeremy Bernstein]

I had to run chkdisk this morning again, after having SF 008 play during the night in Arena. No blue screen event yet though

To the programmers here, does any of the following make sense to you regarding what the cause may be? Is it in SF 008 or Arena or in a combination of them?

damaged attribute post (48, "") - [this one seems to be recurring every time this happens]

damaged file post segment 150176

name giving error 121764

restoring green2.png and GREEN24.png


There were some more, but these were the only ones I managed to get written down...

Something to check: as Stockfish works, please examine the amount of memory being used in the Task Manager. Is it constantly (slowly) increasing? If it happens, does it only happen with Persistent Hash enabled, or all the time?

I'm balls-to-the-wall busy for the next two weeks, so I won't have any chance to look at this myself, but I am interested in anything "strange" that you observe about memory usage.

Thanks, Jeremy

[BorgKing]

OK, I'll do that in a day or three.

Btw, this seems to happen after very long games 3-6 hours for 30-40 moves. A few days ago I ran it with 2h each for 30 moves without incidents - (though most of those were won within 25 moves).

(Unfortunately, it and the others really need the extra hours or else they trade rooks into an ultimately dead-lost different coloured bishops ending.)

[BorgKing]

I did this for about an hour. Use Persistent hash box checked but "0" as book depth.

It does indeed seem to rise slowly, in the "private workmemory"

Move 1
1-6 min 546 236kb
after 7min 546 244
after 9min 546 260
after 10min 546 332

Move 2
2min 546 336
8min 546 352
12min 546 476
37min 546 488

as comparison Houdini 3
Move 1
1min 559 692
4min 559 712

Move 2
2min 559 716
4min 559 720

But the most surprising thing to me was that H3 was using 99% of the cpu while SF 008 only had 48-50%

I probably assumed that in SF 008 it would find both processors automatically, or else forgot to change it...

Anyway, after now using 99% cpu
the numbers for "private workmemory" dropped considerably:

1min 56 212 kb
5min 56 284
10min 56 424
15min 56 448

While H3 is still at 559 720