Page 1 of 1

Stockfish 2.1

Posted: Thu May 05, 2011 6:58 am
by noctiferus

Re: Stockfish 2.1

Posted: Mon May 09, 2011 7:56 am
by noctiferus
At the same location is now available a faster recompile (stockfish 2.1.1).

PS-Oops: seems to be some trouble in Linux recompile... Wait!

Re: Stockfish 2.1

Posted: Mon May 09, 2011 9:38 am
by Uly
I skipped Stockfish 2.0 for these reasons, will wait for stable version ;)

And, I suspect it'll be as unusable for analysis as 2.0.1 was, so I'll stick with the Gran2k/flfh series for a while :)

Still, it's great to see improvement, Stockfish has continued to climb my core engines list since version 1.5, I wouldn't be surprised if a future version was my top main engine.

Re: Stockfish 2.1

Posted: Mon May 09, 2011 12:51 pm
by noctiferus
Hi,Uly.
I'm looking at Martin's tournament. Chessbomb, for evaluating positions, moved to sf.2.1 from 2.0.1variations, based on what was suggested in this forum, and evals seem more stable...

Re: Stockfish 2.1

Posted: Tue May 10, 2011 12:43 pm
by noctiferus
jim Ablett fixed a bug in win 64 compile:

http://dl.dropbox.com/u/5047625/stockfish-211-64-ja.zip

Re: Stockfish 2.1

Posted: Tue May 10, 2011 2:59 pm
by noctiferus
Final complete recompile by Jim Ablett, all bugs fixed:

http://www.mediafire.com/?e8op18qiaam76sd


Link to his post:
http://talkchess.com/forum/viewtopic.ph ... d26be8fd3b

Re: Stockfish 2.1

Posted: Wed May 11, 2011 5:39 pm
by alfons
PS-Oops: seems to be some trouble in Linux recompile... Wait!
I do not understand the need for this actually: generic compiled stockfish on both boxes (Intel/AMD in this case) are already faster (than accompanied binary); adding flags '-march=native' or simply '-msse3' and so on give even more convincing results (PGO-compiling with gcc seems intersting too (doesn't matter whether g++ 4.4/4.5/4.6 is used-all deliver fast binaries)).

Re: Stockfish 2.1

Posted: Wed May 11, 2011 11:35 pm
by noctiferus
IMO it would better if you could send your comment directly to Jim Ablett on Talkchess...

( sorry for the PM I sent you: mouse slip..)

Re: Stockfish 2.1

Posted: Thu May 12, 2011 5:07 pm
by alfons
ok - I got icc Version 12.0.4 installed and tried various (and default) switches and settings (incl. pgo) on a (mid june 2010) mac mini (wizz se Debian-linux* of course): It's definitely not worth the hassle, as no reliable speedups emerged of its use opposed to gcc.

Removing this (locally installed) intel product therefore.

*unstable/recurrent results
$stockfish bench 32 1 10 default depth
vary from 835 kn/s (stable; ja-64-bit-binary)
to 855 kn/s (stable, gcc), 965 kn/s (unstable, gcc) and 840 kn/s (stable, icc, pgo), resp. 930 kn/s(unstable->740kn/s also occurs, icc '-axSSE3')