November 9, 2015

Android UCI engine update: Stockfish Dev 2015-10-28 JA

If i call this version 6.99, this would be nice joke hiding some reality in it. For sure, we are close enough to conclude TCEC-8 superfinal and if the next release of Stockfish is entitled v7.0, it will not be too surprising.

I'm sure a lot of ELO is added over Stockfish 6 but i still prefer to wait for v7 before replacing v6 in Rapidroid.

I've just checked these two compiles. Unfortunately, once again arm7 didn't load while arm5 works fine. However, it seems this issue applies to Stockfish only as arm7 versions of Arasan and Maverick are working on my Rockchip 3188 quad core.

There should be something specific in Stockfish source that leads to an incompatibility and it should be something that Arasan and Maverick don't include.

If someone manages to get the arm7 compile working with Chess for Android or Droidfish, please let me know.

Download arm5 binary: HERE

Download arm7 binary: HERE

18 comments:

  1. Engine for ARM7 devices as usual is not working. The usual nonsense message: "Engine shut down."(stopped). :-(

    ReplyDelete
  2. again mate,like the previous 08/10/15 arm7 compile, this arm7 compile doesnt work under chess for android and droidfish:engine error and terminated,why again? as a sf fan im so tired to wait for a new android compile, only peter could help but...,sf team they dont care for android.

    ReplyDelete
  3. Patience... You need to use arm5 and wait for official compile expected after TCEC.

    ReplyDelete
  4. Hi all,

    Sorry Arm 7 build is not working. I have ordered an Arm7 compatible tablet for testing so I won't release non-working builds in future.
    In the meantime here is a new arm7 build of latest Stockfish src. I would be grateful for feedback as to whether it runs for you.

    http://tinyurl.com/omf6h4y

    thanks,
    Jim.

    ReplyDelete
  5. Your last arm7 build runs ok for me. Options, nps, tb, all is right.

    ReplyDelete
  6. Nice to read a good news. I'll check tonite for myself too.

    ReplyDelete
  7. Thanks. I'm on the right track now I think. What about speed? is it faster for you
    than an arm 5 build ?

    Jim.

    ReplyDelete
  8. Speed is the same as that of Peter Osterlund's last public compile (arm7, 100915).

    ReplyDelete
  9. That'll do :-)
    Many thanks for taking the time to test.

    Jim.

    ReplyDelete
  10. Pls I have been searching all over the net for komodo 9.2 chess engine that I can use on chess analyze this. Pls can u share the link?

    ReplyDelete
    Replies
    1. @Dapo: to use Komodo 9.2 you absolutely need to buy it. There is no apk to find somewhere. Besides It's not expensive at all.

      Delete
  11. Jim, on my octa-core devices, NO engine version works at all. What can I do?.

    ReplyDelete
  12. This comment has been removed by the author.

    ReplyDelete
  13. @ Amador Cuesta ..

    What is cpu in your device?
    Let me know and I will compile a Stockfish test build which targets your cpu.
    I think I will need to compile with Neon instructions as well.

    Jim.

    ReplyDelete
  14. The new octa-core handheld devices come with big.LITTLE architecture in TWO separated groups of four chips each. But both groups work separately, not together at the same time. And that is a big problem.
    My new device is the last Samsung tablet, Galaxy Tab S2, with the Exynos 7 Octa 5433 SoC.
    DroidFish 1.58 will include a direct core chosing option from the engine setting instead from the general settingd (as memory or timings), but the problem is the same, as all those cores are supposed to be identical,as in the MP PCs, BUT in the case of big.LITTLE architectures, they ARE NOT.

    ReplyDelete
  15. The new octa-core handheld devices come with big.LITTLE architecture in TWO separated groups of four chips each. But both groups work separately, not together at the same time. And that is a big problem.
    My new device is the last Samsung tablet, Galaxy Tab S2, with the Exynos 7 Octa 5433 SoC.
    DroidFish 1.58 will include a direct core chosing option from the engine setting instead from the general settingd (as memory or timings), but the problem is the same, as all those cores are supposed to be identical,as in the MP PCs, BUT in the case of big.LITTLE architectures, they ARE NOT.

    ReplyDelete
  16. I've been posting similar comments about big.little cpus. Yes they are not true 8 simultaneous cores.
    But make sure you practice two tests, one with 4 threads and another with 8 threads setting. Are they close in terms of Nps?

    ReplyDelete
    Replies
    1. And another cool finding: Mesiatek 6595 is very suitable for chess. Maybe the best compatibility & perfo & cost combination. It has little. bIG but can activate both at the same time! Plus it's the last mohican of 32 bit cpus to run all existing engines...

      Delete