Stockfish 2.3.1 update available

[Modern Times]

I think that the conclusion is not to use that gui.
I do not see that black evaluated the position as loss for itself so there is no logical reason for a GUI to decide that stockfish won the game.

Or turn off all adjudication.

[bnemias]

1) For people compiling themselves the Stockfish bench signature is: 5423738
It means that if you run ./stockfish bench then it should show 5423738 nodes searched.
This means that the compile is correct.

I just wanted to make a quick comment on this. The bench "signature" does depend on any extra arguments sent. Due to hardware limitations, I had to run

Code: Select all

./stockfish bench 32

And got a different number. But comparing against your build (but using 32 Hash), the numbers matched. So I am assuming everything is fine.

[mcostalba]

So I am assuming everything is fine.

Yes it is fine. What hardware are you using?

[bnemias]

So I am assuming everything is fine.

Yes it is fine. What hardware are you using?

An embedded device without 128MB free ram. It's interesting that the default hash size is 32MB but the "bench" cmdline uses 128MB. It doesn't actually bother me (I have to work around small issues constantly due to running on embedded devices), but I'm curious why bench doesn't just use the regular hash size.

[mcostalba]

but I'm curious why bench doesn't just use the regular hash size.

No reason, so I have just pushed a patch to set deault 'bench' TT size to 32 MB.

What other issues/warnings have you found compiling SF?

[ZirconiumX]

Code: Select all

g++ -g -Wall -Wcast-qual -fno-exceptions -fno-rtti  -ansi -pedantic -Wno-long-long -Wextra -Wshadow -arch ppc -DNDEBUG -O3 -DNO_PREFETCH   -c -o material.o material.cpp
endgame.h: In member function 'MaterialEntry* MaterialTable::probe(const Position&)':
endgame.h:117: warning: 'sf' is used uninitialized in this function

Matthew:out

[mcostalba]

This is a compiler bug, you are using an old version of gcc. I knew the warning and it disappeared with newer gcc.

It is due to calling function map() in:

Code: Select all

  template<typename T> T probe(Key key, T& eg)
  { return eg = map(eg).count(key) ? map(eg)[key] : NULL; }

where map() is defined as:

M1& map(M1::mapped_type) { return m1; }
M2& map(M2::mapped_type) { return m2; }

The warning is due to call map(eg) and because eg could be uninitialized gcc warns. What old gcc misses is that in map() only the type of 'sf' is used to discriminate between the 2 overloads of map(), so the value in itself has no use, indeed as you can see in map() definition there is only the argument type not the variable name. IOW if map would have been defined as:

Code: Select all

M1& map(M1::mapped_type sf) { return m1; }
M2& map(M2::mapped_type sf) { return m2; }

then gcc would be right.....but is not

[ZirconiumX]

Another thing - the patch 'Improve compatibility with older versions of Mac OS X' actually does the exact opposite of what the patch author intended it to do.

-mmacosx-version-min=10.6 prevents the code from running on anything lower than 10.6. I have Mac OS X 10.4. 10.6 is higher than 10.4 so Stockfish won't run. It is as if you are deprecating PowerPC entirely - and not supporting 10.5 on Intel. (Some applications won't support 10.6 because of issues with the APIs etc...)

Matthew:out

[mcostalba]

Yes, thanks. I have seen you opened the issue. Probably Daylen will take care of it. Mac platform maintenance is usually up to him that owns the hardware.

[ZirconiumX]

I could take up the reins for PowerPC if you want?

Matthew:out