CFish

[syzygy]

One note: 12.9. version gives in bench 6024713 at first run, then totally different values like 5122529 4862840 etc. .

Ah, it seems SF clears hash etc. between runs. Thanks

So did Cfish, but the counter move history tables were not correctly being cleared. Fixed now.

[Ras]

I'd be interested in benchmark numbers. Is CFish faster than a functionally identical Stockfish? If so, by how much?

[Dann Corbit]

I'd be interested in benchmark numbers. Is CFish faster than a functionally identical Stockfish? If so, by how much?

It is shockingly faster on my hardware (nearly twice as fast from the middle game onwards).

I suspect that Ronald has done more than a literal translation and made a few improvements like Mohammed did with Asmfish.

[Eelco de Groot]

I'd be interested in benchmark numbers. Is CFish faster than a functionally identical Stockfish? If so, by how much?

It is shockingly faster on my hardware (nearly twice as fast from the middle game onwards).

I suspect that Ronald has done more than a literal translation and made a few improvements like Mohammed did with Asmfish.

I don't know about the referenced speeds from Dann, but the executables with GCC can be very small indeed. The NPS I got when compiling with TDM-GCC are roughly the same as with Stockfish compiled here, maybe a bit higher though.

If anyone is interested in a Ferrarifish III version of Cfish, I made that and called it Dolphin, as I had not seen a Stockfish clone with that name and it really is a nice name for a Stockfish based derivative chess program although technically it is not really a fish. No more than 7 lines changed in Ronald's ntsearch.c

It is available on the Toga Developers Bulletin Board.

Dolphin is only 183 kB! Cfish is 186 kB. I hope it can run on other systems though, if not I can't really make a 64 bit compile for other systems than my own. Also no 32 bit compile, I am not sure Cfish supports that.


Rainbow Dolphin


No crashes with cfish master, I played a very short 2 min per game per player match against Stockfish:

Cfish_001 2016-09-17 - Sf Master 2016-09-14: 5 - 5 (+0 =10 -0) 50% -0 Elo according to Shredder

[Damir]

Hi Eelco

Can you make other compile than BMI2 ? My system don't support it.

[Rochester]

The Dolphin chess is the copyright of Mr Nguyễn Việt Huy, yes? Good he can be back. GUI is beauty!

[Eelco de Groot]

Hi Eelco

Can you make other compile than BMI2 ? My system don't support it.

Hi Damir,

Yes, at least, I hope so, I edited the post.

Thanks to the 8 downloaders so far !

[Eelco de Groot]

The Dolphin chess is the copyright of Mr Nguyễn Việt Huy, yes? Good he can be back. GUI is beauty!

Sorry Roch, I did not know of a program called Dolphin Chess, but anyway this is just a Cfish MOD, I hope no offense to the programmer of Dolphin Chess.

[Damir]

Thanks Eelco. I have downloaded your engine now.

[Ras]

It is shockingly faster on my hardware (nearly twice as fast from the middle game onwards).

Wow, that is impressive. In identical positions, are the calculated PVs the same?

I suspect that Ronald has done more than a literal translation and made a few improvements like Mohammed did with Asmfish.

Even then - the Stockfish developers are already top-league ones so that doubling the speed with just some small changes seems improbable. The Stockfish team would have harvested these major gains long ago.

You see, there is much discussion whether C++ is slower than C or not; the usual argument by C++-developers is that C++ is on a par with C, provided that you have very competent C++ developers since it is much easier to go wrong with C++. Only that there is a shortage on complex applications that would allow for a direct 1:1 comparison.