Smaug: a new chess engine based on glaurung

[mcostalba]

Sure I'm interested of it! Thanks! You can post it to jo3on5a.k7iis6ki@gm8ail.com (remove numbers!). Perhaps I can avoid reinventing the wheel

I have sent to you an email right now.

Regarding the test of eval tweaks we are now at 749 games out of 1000 and current result is +195 -173 = 381 +10 ELO

As said I have not tested the second patch but after quick running it on a set of positions I have see that with the patch applied the number of searched nodes is 20% less at fixed depth!!!!

If the patch is so good that avoids to erroneusly prune important moves then I foreseen a great result also on the second one.

to be continued....

Marco

[Denis P. Mendoza]

While waiting for Jim's faster ones, you can try these ICC11 Windows builds:

http://www.mediafire.com/download.php?qym0gmtjat2

[mcostalba]

Ok I have finised the test with the evaluation parameters tweaks:

Engines: Stockfish with patch vs Stockfish without patch

Result after 999 1+0 games: +264 -230 =505 +12 ELO

So patch is VERY good and has been eagerly merged in Stockfish development version.

Now testing null move beta corrections.....

Thanks
Marco

P.S: BTW Joona, from where do you got the idea of NULL move beta corrections? I had a similar idea (but much less tought out, with a fixed delta and not so refined and I gave up after few tests).

[sje]

The Intel C++ compiler is available for a limited time free trial if you give them some marketing data, Otherwise it's US$599 plus US$240/year for support.

I tried the free version some time ago. I don't recall any earthshaking improvements over g++, but I do remember that the Intel compiler did supply a few more useful diagnostics about questionable coding.

[zamar]

While waiting for Jim's faster ones, you can try these ICC11 Windows builds:

http://www.mediafire.com/download.php?qym0gmtjat2

Thank you for the builds. In few days, I'll put Mac- and Windows-builds on my homepage.

[zamar]

Ok I have finised the test with the evaluation parameters tweaks:

Engines: Stockfish with patch vs Stockfish without patch

Result after 999 1+0 games: +264 -230 =505 +12 ELO

So patch is VERY good and has been eagerly merged in Stockfish development version.

The result is a bit better I expected. It's also final confirmation that my over-simplified optimization method works. I'm more than happy.

Now testing null move beta corrections.....

Interesting to see how well those values apply to Stockfish...

P.S: BTW Joona, from where do you got the idea of NULL move beta corrections? I had a similar idea (but much less tought out, with a fixed delta and not so refined and I gave up after few tests).

Correction/Delta/Tempovalue is mentioned many papers related to NullMove heuristic. Scaling it using game phase and search depth is however (as far as I know) a new idea. Depth affects a lot to the nature of search through different prunings and reductions plus it's widely known that Null move heuristic works well in (often tactical) midgame positions and quite badly in endgame.

[zullil]

The Intel C++ compiler is available for a limited time free trial if you give them some marketing data, Otherwise it's US$599 plus US$240/year for support.

I tried the free version some time ago. I don't recall any earthshaking improvements over g++, but I do remember that the Intel compiler did supply a few more useful diagnostics about questionable coding.

The Intel C++ compiler for Mac OS X is available for 30 days at no charge, for evaluation. No submission of marketing data is required.

Chess engine binaries compiled with icc are almost twice as fast as those compiled with gcc, at least on my Mac. For example, I just compiled Smaug221 using both icpc and g++. In a middlegame position with 45 possible moves, the icpc binary examined 1.81M nps, while the g++ binary examined 0.92 M nps.

Your results may vary a bit, but the Intel compiler produces much faster binaries than gcc does.

[krazyken]

The Intel C++ compiler is available for a limited time free trial if you give them some marketing data, Otherwise it's US$599 plus US$240/year for support.

I tried the free version some time ago. I don't recall any earthshaking improvements over g++, but I do remember that the Intel compiler did supply a few more useful diagnostics about questionable coding.

The Intel C++ compiler for Mac OS X is available for 30 days at no charge, for evaluation. No submission of marketing data is required.

Chess engine binaries compiled with icc are almost twice as fast as those compiled with gcc, at least on my Mac. For example, I just compiled Smaug221 using both icpc and g++. In a middlegame position with 45 possible moves, the icpc binary examined 1.81M nps, while the g++ binary examined 0.92 M nps.

Your results may vary a bit, but the Intel compiler produces much faster binaries than gcc does.

That seems a bit odd. What flags did you use with gcc? With my builds when I tried out ICC the ICC builds were rarely more than 10% faster. My first guess would be that you didn't compile 64 bit with gcc.

My default flags for compiling with gcc include:

Code: Select all

-O3 -fast -m64 -mmmx -msse3

[sje]

The Intel C++ compiler is available for a limited time free trial if you give them some marketing data, Otherwise it's US$599 plus US$240/year for support.

I tried the free version some time ago. I don't recall any earthshaking improvements over g++, but I do remember that the Intel compiler did supply a few more useful diagnostics about questionable coding.

The Intel C++ compiler for Mac OS X is available for 30 days at no charge, for evaluation. No submission of marketing data is required.

Chess engine binaries compiled with icc are almost twice as fast as those compiled with gcc, at least on my Mac. For example, I just compiled Smaug221 using both icpc and g++. In a middlegame position with 45 possible moves, the icpc binary examined 1.81M nps, while the g++ binary examined 0.92 M nps.

Your results may vary a bit, but the Intel compiler produces much faster binaries than gcc does.

As I wrote, I never saw any huge speed increase in spite of considerable experimentation with compiler options. The extra diagnostics were useful even if half of them were spurious to some extent.

Also: Apple is spending a lot of R&D budget on speeding up Mac OS/X and the various professional applications it sells. Yet all of these are built with the same gcc tools that Apple ships for free. If the Intel compiler were really that much better, then Apple would use it themselves.

[zullil]

The Intel C++ compiler is available for a limited time free trial if you give them some marketing data, Otherwise it's US$599 plus US$240/year for support.

I tried the free version some time ago. I don't recall any earthshaking improvements over g++, but I do remember that the Intel compiler did supply a few more useful diagnostics about questionable coding.

The Intel C++ compiler for Mac OS X is available for 30 days at no charge, for evaluation. No submission of marketing data is required.

Chess engine binaries compiled with icc are almost twice as fast as those compiled with gcc, at least on my Mac. For example, I just compiled Smaug221 using both icpc and g++. In a middlegame position with 45 possible moves, the icpc binary examined 1.81M nps, while the g++ binary examined 0.92 M nps.

Your results may vary a bit, but the Intel compiler produces much faster binaries than gcc does.

That seems a bit odd. What flags did you use with gcc? With my builds when I tried out ICC the ICC builds were rarely more than 10% faster. My first guess would be that you didn't compile 64 bit with gcc.

My default flags for compiling with gcc include:

Code: Select all

-O3 -fast -m64 -mmmx -msse3

Yup, my mistakes. When I recompiled using g++-4.2 and included -m64 I got a binary that examined 1.64M nps. So icc is about 10% faster.

Thanks for correcting my error.