Page 1 of 5

New engine: Xiphos

Posted: Wed Feb 28, 2018 7:30 am
by mt
I released an initial version of my new UCI chess engine.

Github: https://github.com/milostatarevic/xiphos
Binaries: https://github.com/milostatarevic/xiphos/releases

Please note that the project is still work in progress. The evaluation function is relatively simple (~200 lines of code), and together with some other parts, can be further improved.

Regarding the strength, it should be on par with engines rated around 3000 Elo on CCRL 40/4 scale. I compared it only to a few other engines, so this could be off.

The linux/osx bmi2 versions are extensively tested, windows not that much, so please tell me if there are any issues. When tested locally, bmi2 version was about 12 Elo points ahead of the version that implemented the magic bitboards. This might not be true for some other CPUs. If in doubt which version to use, please check the nps (for example, by running perft).

I hope you'll find my project interesting.

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 8:17 am
by BrendanJNorman
mt wrote:I released an initial version of my new UCI chess engine.

Github: https://github.com/milostatarevic/xiphos
Binaries: https://github.com/milostatarevic/xiphos/releases

Please note that the project is still work in progress. The evaluation function is relatively simple (~200 lines of code), and together with some other parts, can be further improved.

Regarding the strength, it should be on par with engines rated around 3000 Elo on CCRL 40/4 scale. I compared it only to a few other engines, so this could be off.

The linux/osx bmi2 versions are extensively tested, windows not that much, so please tell me if there are any issues. When tested locally, bmi2 version was about 12 Elo points ahead of the version that implemented the magic bitboards. This might not be true for some other CPUs. If in doubt which version to use, please check the nps (for example, by running perft).

I hope you'll find my project interesting.
Pardon my asking, but is this a brand new original engine, or is it based on existing engines (Stockfish, Ipo etc?).

The reason I ask is that it is really strong, and from what I've seen in a few games so far, it plays beautiful chess too.


BTW...here's a logo:

Image :wink:

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 8:33 am
by mt
Hi Brendan, thanks. Yes, the engine is written from scratch.

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 9:32 am
by Damir
Thanks a lot for the new engine Milos. I hope you continue with your project, who btw sound very interesting. :) :)

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 9:41 am
by LocutusOfPenguin
Hi Milos,

i had to change the makefile to include threads (p not l) and math (at end) like so:
gcc -pthread -O3 -flto -mbmi2 -Wall src/*.c -o xiphos -lm


Could y also implement the "go searchmoves" please?

Jürgen

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 9:47 am
by sandermvdb
An initial version that is ~3000 elo! :o
How long have you been working on this project?

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 10:17 am
by Damir
Milos, I am having problems when trying and load the engine on Playchess server. It will not load. I am using BMI2 version of your engine. I hope you can fix the problem. :? :?

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 12:44 pm
by velmarin
Congrats, is a very promising project, although I haven't been able to prove it.
Windows_x64_magic version does not work on my SSE2 machine

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 2:13 pm
by Dokterchen
Awesome! Thanks Milos

Re: New engine: Xiphos

Posted: Wed Feb 28, 2018 3:13 pm
by CMCanavessi
Cool, a nice engine for Season 3 of my tournament! Let's see how it does.