The most stupid idea by the Stockfish Team

[Gabor Szots]

@Gabor
It seems your msys2 installation never was really complete, as grep is already in the 'base' group.
https://packages.msys2.org/group/

I suggest running all those commands again one by one, it will only load, what is missing or otherwise update.

Code: Select all

pacman -Syu
pacman -S mingw-w64-x86_64-gcc
pacman -S mingw-w64-x86_64-make
pacman -S mingw-w64-x86_64-toolchain
pacman -S libraries

If you ever want to compile e.g. cutechess you need this too

Code: Select all

pacman -S mingw-w64-x86_64-qt5-static

Hi Günther,

Thanks for the hints. I did a complete -Syu and -Su this morning then did everything as you had told me, no success. I'm going to investigate further, as a last resort I may delete the whole msys2 and reinstall everything again.

BTW, pacman -Si did list grep.

[Gabor Szots]

To Ronald and Günther:

While I certainly don't understand what I am doing, the result is success. I have installed manually the whole Cygwin base package therefore grep as well. Now CFish compiles perfectly. Thank you both for yor help.

What I still don't understand, although it is not important any more, is why I have this directory in my MSYS2 installation:

Directory of e:\msys64\var\lib\pacman\local\grep-3.0-2

2020.09.01. 16:42 <DIR> .
2020.09.01. 16:42 <DIR> ..
2020.09.01. 13:44 354 desc
2020.09.01. 13:44 4˙334 files
2020.09.01. 16:42 0 grep.txt
2018.12.05. 12:53 326 install
2018.12.05. 12:53 4˙271 mtree
5 File(s) 9˙285 bytes

[syzygy]

Is CFish now at the same commit level again as SF dev currently?

It is a couple of days behind at the moment.

It seems it is faster than my SFdev compile w/o nnue included. Need to check this more precisely.

The NNUE version should also be faster, in particular if you have an older CPU.

[Dann Corbit]

If I could get Cfish to log analysis to disk, I would use it for most of my chess analysis.

I could just make a C interface to the C++ file I use for logging with all of the SF variants, but I won't do that because I feel like it is contrary to the spirit of Cfish (minimalistic C approach using the same algorithms as Stockfish).

[Guenther]

Is CFish now at the same commit level again as SF dev currently?

It is a couple of days behind at the moment.

It seems it is faster than my SFdev compile w/o nnue included. Need to check this more precisely.

The NNUE version should also be faster, in particular if you have an older CPU.

That's what I meant (w/o nnue included = nnue NOT embedded, but used from outside).
Your CFish is a lot faster than SF Devs, compiled by me.
Yours is around 400kn/s, SF dev nnue embedded around 290 and nnue not embedded around 300.
Measured at the date of my previous post from start position, may be depth 25-27.

[Dann Corbit]

Is CFish now at the same commit level again as SF dev currently?

It is a couple of days behind at the moment.

It seems it is faster than my SFdev compile w/o nnue included. Need to check this more precisely.

The NNUE version should also be faster, in particular if you have an older CPU.

That's what I meant (w/o nnue included = nnue NOT embedded, but used from outside).
Your CFish is a lot faster than SF Devs, compiled by me.
Yours is around 400kn/s, SF dev nnue embedded around 290 and nnue not embedded around 300.
Measured at the date of my previous post from start position, may be depth 25-27.

CFish on my AMD 3970x absolutely rocks.
And AVX2 added so much speed, people will think I am lying if I say it.

[Dann Corbit]

If you have a recent AMD and don't want to build it yourself: