Fastest SF8 compile?

[IWB]

Hello all

i am interested to see how much Assembler is really worth compared to an optimized C-compile. Therefore I am looking for the fastest SF8 compile (original release code). I am only interested in a real SF8 C-compile, not ASM fish.
Anyone with a link for me?

Thanks
Ingo

[gogamoga]

You can use SF BYO to compile your own executable specifically for your machine

[IWB]

That doesn't work for my purpose because someone could optimise the asm code as well ...
I need a (few) general purpose SF8 C compiles with POPCNT. I like to check them on my hardware and will choose the fastest one.

Thanks anyhow
Ingo

[pferd]

I compiled the bmi2 version of sf_8 on my own and the binary provided on stockfishchess.org was faster than mine. I think the same applies to the modern version. I do not think there is much room left for improvement. Kiran did a pretty good job back then.

[pohl4711]

I compiled the bmi2 version of sf_8 on my own and the binary provided on stockfishchess.org was faster than mine. I think the same applies to the modern version. I do not think there is much room left for improvement. Kiran did a pretty good job back then.

I think so, too. Normally I use the pocount-compiles of I.Ivec (ultimaiq-compiles) http://chess.ultimaiq.net/stockfish.html for my Stockfish-Bullet testruns. For the Stockfish 8 testrun, I used the official popcount compile from stockfishchess.org. No big difference in nps as far as I remember. The official SF8-compiles from stockfishchess.org are pretty fast.

[Jouni]

Fastest I have tried is Kiran Panditrao's exe with additional dll-files. But I prefer one file version more.

[IWB]

Hello all

Thanks for your your proposals and help. I tried a couple of different compiles and the original one seems to be quite good!

It seems that the difference between a modern C++ compiler and an ASM version is worth about 20 Elo. My personal feeling is that this is respectable but not revolutionary.

Thanks again
Ingo