HAL9000

HAL9000
"It just isn't conceivable that you can design a program strong enough to beat players like me."

July 11, 2015

Strongest Stockfish 6 compile for Android?

Now that i'm done with the multiversion lizard-o-fishy battle, it wouldn't be time wasting to come back to the previous dilemma about different Stockfish 6 compiles.

Just a couple of weeks after TCEC-7, Stockfish 6 had been released step by step, first with release candidates and the definitive version had followed soon after.

Although the original compile for Android didn't delay, other alternatives rained from different sources to complete the chaotic complexity:
* RC6 arm5 by Jim Ablett
* 6.0 arm5 by Jim Ablett
* 6.0 arm5 bundled with Droidfish 1.57
* 6.0 arm7 bundled with Droidfish 1.57
* 6.0 arm8 bundled with Droidfish 1.57 for 64 bit Android Lollipop
* 6.0 original arm7 by stockfishchess.org

Which is the strongest of these? Good question. It's almost certain arm7 is better than arm5 in same conditions unless big difference between compilers.

Basically Jim Ablett compiles performed very well in the past, especially in Senpai 1.0 case but there are few chances for his compiles to overcome Peter Ă–sterlund's.

What i really wonder is the difference between Droidfish and the original one. As i can't live my life without knowing about this, i'm starting a new parallel experiment between all above compiles with the exception of arm v8 version. I don't have Lollipop and probably i will never go for it because almost all existing compiles will refuse to run under Chess for Android. As mentioned before, no tourneys with Lollipop yet!

Let me remind that the Stockfish compile running in Rapidroid is the one released by stockfishchess.org, made with MingW under Linux.

Test conditions:
* Novatab tablet with quad core Rockchip 3188 downclocked to 1.0 Ghz
* 256MB hash
* No ponder, no tablebases
* TC 1200+2
* TCEC7 openings of 8 moves

First results to come soon.

No comments: