Not unique build process and not same performances about koivisto

[Luecx]

Thank you! Will change it asap

[Luecx]

200 Elo seems very extreme.
Considering that the engine is deterministic, such a elo difference could only be explained by having a 10x speedup.
What did you try to test them against each other?

[Luecx]

I will have a look at it myself real quick

[tryingsomestuff]

I'm using cutechess tourney. Short TC, 10s+0.1.
I'm running a 3s+0.1 now with koivisto_cmake_make and koivisto_native to show you some results.

Nps are indeed of my x5!

Koivisto_native : 2Mnps
Koivisto_cmake_root_dir : 0.4Mnps

Score of Koivisto vs Koivisto_cmake_make: 13 - 0 - 3 [0.906]
...      Koivisto playing White: 8 - 0 - 0  [1.000] 8
...      Koivisto playing Black: 5 - 0 - 3  [0.813] 8
...      White vs Black: 8 - 5 - 3  [0.594] 16
Elo difference: 394.1 +/- nan, LOS: 100.0 %, DrawRatio: 18.8 %
16 of 999999 games finished.

[tryingsomestuff]

Is the default cmake target a debug executable ?

[Luecx]

It is missing optimization flags.
I use CLion and I can choose if the cmake shall be used as a debug or release build.

I assume the default one is a debug build.

Thanks for pointing this out!

[Luecx]

Do you have an idea on how this could we solved?

Would is just help to just add O3 to the cmakelists.txt?

[tryingsomestuff]

Maybe the question is why having 2 build process ?

[tryingsomestuff]

Yes maybe adding

set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3")

might help

[Luecx]

One build-process is used for development using CLion. Although we strongly recommend using the makefile which is far more optimised.

[Luecx]

Have you tested if your proposed change works?

[tryingsomestuff]

yep, looks like same nps and strength follows

[Luecx]

Fixed it. The last thing is that the version number is not correct when building with cmake. But this does not really matter :)