Vajolet 2.1 released

[Frank Quisinsky]

Hi Marco,

your and two other engines gave me the motiviation for starting with my FCT League systems after I am working a longer time on SWCR Rating List. I lost on SWCR my interest for different reasons. Now after more as 2 years FCT League Systems I am full motivated with all the nice developments I can see daily on my PCs.

Since this time I started with FCT League System Vajo played around 10.000 45-minutes games (40/10) on my i7 4.0 Ghz systems and believe me I have here a lot of fun. Version 2.0 made a big jumping to clearly more interesting chess in middlegame. Each time good for a surprise ... no other engine should underestimate Vajo.

Thanks for your new 2.1 version.
Version 2.0 runs without any problems in around 4.000 games.

I will test v2.1 as soon as possible.
At the moment I need more engines for my FCT League 3 systems. Here I have different ideas to make all more interesting. But I need three FCT League systems for it.

Vajo and all other new updates have to wait but you can be sure ... will be test it as soon as possible.

Have a good new year!
Thanks again, I am very happy about the new Vajo.

Best
Frank

[Frank Quisinsky]

Could be possible that you like to read bad english.
In such a case I have a bit for you ...

http://www.amateurschach.de/main/_playing-style.htm

Best
Frank

[elcabesa]

I already read it long time ago :
thank you very much

[tpoppins]

I am impressed with the low branching factor as well.

There are some issues when running the POPCNT exec under Chessbase GUIs, though - tested with the latest Fritz 15 GUI version.

1) THREADS is capped to 8. Is there any reason for it? I'd expect a Lazy SMP engine to use as many as possible.

2) Max hash appears to be capped to 4096, going by the UCI strings in the exec - another illogical limit. Anyway, selecting 4096 in GUI causes the engine to freeze and the GUI spit an exception error. 1024, 2048 and 3072 appear to work.

3) After installing the engine the GUI reports its name as "vajolet2 2.1x64 pop". This looks a little untidy (run-on words, no space between 2.1 and x64) though still acceptable. However, Chessbase's Let's Check function cannot read the engine name and reports it as "New Engine". This could be caused by non-standard formatting or location of the engine strings in the exec (just a guess on my part, I have little knowledge of UCI specs).

The last issue could be blamed on Chessbase, but the vast majority of engines I've used, including dozens of SF dev builds, several Komodo versions going back to 5.1r, lesser engines like Gull and Fire and half a dozen various builds of Crafty 25.0, don't exhibit this problem. Only a handful engines - SugarPro, DeepFishMZ and Laser - show up as "New Engine" in Let's Check, which leads me to think the problem is with the authors not following certain conventions.

All of these are minor issues and should be easy to fix quickly. Looking forward to using this interesting engine more extensively.

[Frank Quisinsky]

Hi Marco,

selection will be overwork with a new idea later this year. Since a while I am working with full power on my own opening book and all other things have lesser priority.

Now Vajolet runs on my Notebook ...
I am crazy ...

Time for bughouse, hope that I can used here all my PCs.



Best
Frank

[elcabesa]

1) THREADS is capped to 8. Is there any reason for it? I'd expect a Lazy SMP engine to use as many as possible.

there is no reason for not raise the number of Treads, I think I'll raise the threads in the next release.

2) Max hash appears to be capped to 4096, going by the UCI strings in the exec - another illogical limit. Anyway, selecting 4096 in GUI causes the engine to freeze and the GUI spit an exception error. 1024, 2048 and 3072 appear to work.

it's indeed a bug. it work up to 4095. I'll fix in th next release. I'll probably try to reaie the limits.

No one has ever bothered about those limits most of the tournaments were played with lower hash and threads. I'll try to raise the limits.

3) After installing the engine the GUI reports its name as "vajolet2 2.1x64 pop". This looks a little untidy (run-on words, no space between 2.1 and x64) though still acceptable. However, Chessbase's Let's Check function cannot read the engine name and reports it as "New Engine". This could be caused by non-standard formatting or location of the engine strings in the exec (just a guess on my part, I have little knowledge of UCI specs).

I tried to use the engine with chessbase reader, it simply work and find the name of the engine. I don't know how to help you

[tpoppins]

Thank you for looking at these minor issues, Marco.

I think it would be a good idea to raise the threads and hash limits to the max allowed by the x64 architecture (or x64 Windoze, at least ;). In a couple of years 128-core/1TB RAM servers may become commonplace, no reason why your engine should not be able to extract the max power from the future hardware.

As for the engine name problem, it only shows up in Let's Check. That feature is only available in Chessbase's commercial releases (IIRC, Deep Fritz 13 and later, CB 12 and later), so there's probably no way you could test it with the freeware CBReader (you'd also need a valid Playchess account for the system to accept your analysis).

The only thing I can think of is comparing your sources with those of engines that Let's Check recognizes (e.g. Stockfish) and trying to eliminate any differences in formatting and such. It might take a few test builds and I'd be happy to test them for you.

Thank you for your time and effort, Marco.

[cdani]

Congratulations!

[Jamal Bubker]

Grazie Mille Marco !!

[Max]

I'm pleased to release a new release of Vajolet2

this is my first version with lazy smp code. despite lot of works and debugging the new version is probably ony few Elo point stronger.

in the zip you could find a 64 bit executable, a 32 bit executable and old 64bit version compiled for Gabor Szots


you can download it at https://drive.google.com/file/d/0B8vVDp ... sp=sharing

Hi Marco,

had no luck compiling Vajolet2 2.1 on my Mac. Do you have some hints?

Did "g++ *.cpp -o vajolet21 -O3 -std=c++11 -fmessage-length=0 -mpopcnt" with gcc 4.9 and got this error:

Code: Select all

search.cpp: In member function 'Score search::startThinking(searchLimits&)':
search.cpp&#58;311&#58;170&#58; error&#58; no matching function for call to 'std&#58;&#58;thread&#58;&#58;thread&#40;<unresolved overloaded function type>, __gnu_cxx&#58;&#58;__alloc_traits<std&#58;&#58;allocator<search> >&#58;&#58;value_type*, int, unsigned int, Score&, Score&, std&#58;&#58;reference_wrapper<std&#58;&#58;list<Move> >)'
      helperThread.push_back&#40;std&#58;&#58;thread&#40;alphaBeta<search&#58;&#58;nodeType&#58;&#58;HELPER_ROOT_NODE>,&helperSearch&#91;i&#93;,0,&#40;depth-reduction+(&#40;i+1&#41;%2&#41;)*ONE_PLY,alpha,beta,std&#58;&#58;ref&#40;pvl2&#91;i&#93;)));
                                                                                                                                                                          ^
search.cpp&#58;311&#58;170&#58; note&#58; candidates are&#58;
In file included from thread.h&#58;21&#58;0,
                 from search.cpp&#58;33&#58;
/opt/local/include/gcc49/c++/thread&#58;133&#58;7&#58; note&#58; std&#58;&#58;thread&#58;&#58;thread&#40;_Callable&&, _Args&& ...) &#91;with _Callable = int &#40;search&#58;&#58;*)&#40;unsigned int, int, int, int, std&#58;&#58;list<Move>&); _Args = &#123;search*, int, unsigned int, int&, int&, std&#58;&#58;reference_wrapper<std&#58;&#58;list<Move, std&#58;&#58;allocator<Move> > >&#125;&#93;
       thread&#40;_Callable&& __f, _Args&&... __args&#41;
       ^
/opt/local/include/gcc49/c++/thread&#58;133&#58;7&#58; note&#58;   no known conversion for argument 1 from '<unresolved overloaded function type>' to 'int &#40;search&#58;&#58;*&&)&#40;unsigned int, int, int, int, std&#58;&#58;list<Move>&)'
/opt/local/include/gcc49/c++/thread&#58;128&#58;5&#58; note&#58; std&#58;&#58;thread&#58;&#58;thread&#40;std&#58;&#58;thread&&)
     thread&#40;thread&& __t&#41; noexcept
     ^
/opt/local/include/gcc49/c++/thread&#58;128&#58;5&#58; note&#58;   candidate expects 1 argument, 7 provided
/opt/local/include/gcc49/c++/thread&#58;122&#58;5&#58; note&#58; std&#58;&#58;thread&#58;&#58;thread&#40;)
     thread&#40;) noexcept = default;
     ^
/opt/local/include/gcc49/c++/thread&#58;122&#58;5&#58; note&#58;   candidate expects 0 arguments, 7 provided

-Max