Winboard Forum

I've managed (at long last!) to compile a windows
executable of the Apilchess/Hydra uci chess engine source
code by U. Lorenz and Chrilly Donninger.

http://wwwcs.uni-paderborn.de/fachberei ... chess.html

Primarily designed for custom multiprocessor/parallel computing
hardware, it also works on single processor systems, but is
obviously significantly weaker as a result.

Apilchess's code is very sophisticated, but it's strength (on
single processor systems) is pretty weak, evaluation is
under-developed - about equal or slightly better than TSCP.
It plays a lot better using an opening book, without one, it
always seems to start with Anderssen opening a2-a4!

When I first compiled the code, the hashtable memory allocation
was 350mb! I've adjusted this to 65mb. A different hashtable size
can be compiled on request. I've also made small adjustments to
time allocation and static piece values.

Also there is a folder called 'error' in the ApilChess folder - this
has to be there! and when you first start ApilChess it opens a port
on your computer (not sure why - something to do with the i/o pipes in
MPich?) that will alert your firewall. If you're worried about this
just block it.

Since this early code was released in 2003, the Hydra project has
progressed to the point where it is probably the strongest chess
computer ever - read here > http://www.hydrachess.com/


Intel optimized build.
Windows-adapted source code available on request.


Thanks to Dann Corbit for compiling advice posted on the
old Winboard forum.

Available from my homepage.

Jim.

Jim Ablett wrote:

I've managed (at long last!) to compile a windows
executable of the Apilchess/Hydra uci chess engine source
code by U. Lorenz and Chrilly Donninger.

http://wwwcs.uni-paderborn.de/fachberei ... chess.html

Primarily designed for custom multiprocessor/parallel computing
hardware, it also works on single processor systems, but is
obviously significantly weaker as a result.

Apilchess's code is very sophisticated, but it's strength (on
single processor systems) is pretty weak, evaluation is
under-developed - about equal or slightly better than TSCP.
It plays a lot better using an opening book, without one, it
always seems to start with Anderssen opening a2-a4!

When I first compiled the code, the hashtable memory allocation
was 350mb! I've adjusted this to 65mb. A different hashtable size
can be compiled on request. I've also made small adjustments to
time allocation and static piece values.

Also there is a folder called 'error' in the ApilChess folder - this
has to be there! and when you first start ApilChess it opens a port
on your computer (not sure why - something to do with the i/o pipes in
MPich?) that will alert your firewall. If you're worried about this
just block it.

Since this early code was released in 2003, the Hydra project has
progressed to the point where it is probably the strongest chess
computer ever - read here > http://www.hydrachess.com/


Intel optimized build.
Windows-adapted source code available on request.


Thanks to Dann Corbit for compiling advice posted on the
old Winboard forum.

Available from my homepage.

Jim.

Hi Jim

Cool! Thanks

I?m going to test it.

Best Regards

Jim Ablett wrote:

I've managed (at long last!) to compile a windows
executable of the Apilchess/Hydra uci chess engine source
code by U. Lorenz and Chrilly Donninger. ...

Thanks and congratulation for that. I tried it too (not very hard I admit) but haven't been succesfull.


...

Jim Ablett wrote:Also there is a folder called 'error' in the ApilChess folder - this
has to be there! and when you first start ApilChess it opens a port
on your computer (not sure why - something to do with the i/o pipes in
MPich?) that will alert your firewall. If you're worried about this
just block it.
...

Intel optimized build. ...

Doeas that mean it doesn't run on my AMD? When I try to start it a messages appears complaining about a missing mpich.dll.

Regards


Volker

Volker Pittlik wrote:

Jim Ablett wrote:

I've managed (at long last!) to compile a windows
executable of the Apilchess/Hydra uci chess engine source
code by U. Lorenz and Chrilly Donninger. ...

Thanks and congratulation for that. I tried it too (not very hard I admit) but haven't been succesfull.


...

Jim Ablett wrote:Also there is a folder called 'error' in the ApilChess folder - this
has to be there! and when you first start ApilChess it opens a port
on your computer (not sure why - something to do with the i/o pipes in
MPich?) that will alert your firewall. If you're worried about this
just block it.
...

Intel optimized build. ...

Doeas that mean it doesn't run on my AMD? When I try to start it a messages appears complaining about a missing mpich.dll.

Regards


Volker

Hi Volker

Same problem here in Pentium IV computers and Windows 2000 and XP.

Jim, other question, could you compile it to 16 mb of Hash?

Thanks

Hi Volker,

Sorry, you need to install this package first (forgot about it!)

http://www-unix.mcs.anl.gov/mpi/mpich/mpich-nt/

best,
Jim.

Hi Sergio,

Email me for a 16mb version.

regards,

Jim.

I've checked Apilchess's dll dependencies and all it seems to
require is mpich.dll (full package is needed for multiprocessing)
I've now added the dll to the download package.

I'd be grateful if someone could try it out - see if it works.

Jim.

Jim Ablett wrote:I've checked Apilchess's dll dependencies and all it seems to
require is mpich.dll (full package is needed for multiprocessing)
I've now added the dll to the download package.

I'd be grateful if someone could try it out - see if it works.

Jim.

Thanks Jim,

it seems to work now.

Volker

Michael Diosi reported that Apilchess
wont move unless you have to allow it access when
your firewall pops up. (maybe cluster related?)

Jim.

Jim Ablett wrote:Michael Diosi reported that Apilchess
wont move unless you have to allow it access when
your firewall pops up. (maybe cluster related?)

Jim.

No problems here. I don't see what additional ports are opened (netstat -an). If any I guess that could be to the programs architecture.

Volker

Apilchess's code is very sophisticated, but it's strength (on
single processor systems) is pretty weak, evaluation is
under-developed - about equal or slightly better than TSCP.
It plays a lot better using an opening book, without one, it
always seems to start with Anderssen opening a2-a4!

Which files contain the eval and search?
daniel

Hi Daniel,

apilchess.cpp
chessmaster.cpp

I think?

regards,
Jim.

Update:

1.02 - Some piece value and time control adjustments

Plays a little better.

Available from my homepage


Jim.

Hello Jim

Thanks a lot for this achievement, compiling that one must have been a very tough task

A very quick test : ApilChess always ponder on my PIV (HT on).

Olivier

Hi Olivier,

Thanks a lot for this achievement, compiling that one must have been a very tough task

A very quick test : ApilChess always ponder on my PIV (HT on).

I've been trying to compile Apilchess for about two years on and off.
I was beginning to think it must be some sort of inter-galactic aptitute test set by an alien race.
I'm now waiting for their ship to land in my back garden!

I never noticed Apilchess pondering all the time, I'll see if I can fix it - thanks.

best,

Jim.

New update:

1.03 changes:

1. Added configurable hash memory size.
Add one of these lines to the 'init string'


hash16 or hash32 or hash64 or hash128

Example - Winboard:

(Add this line to winboard.ini)


/firstInitString="hash16" ;


With Arena:

Engines > Manage > Details > Special
and enter it in the 'init string' box

Available from my homepage.

Jim.

Re: Winboard.ini > I mean for Polyglot!

Hello Jim

ApilChess 1.03 crashes on my P4 (Arena 1.1). Here is part of the debug :

Code: Select all

36469*2*Start calc, move no: 1
36469>2:ucinewgame
36469>2:isready
36500<2:readyok
36625>2:position startpos moves e2e4
36625>2:go wtime 1199973 btime 1200000 winc 0 binc 0 movestogo 40
36657<2:info string Userkommando: position startpos


ApilChess is black and the opponent played 1.e4

Olivier

Just tried it on my Athlon64 3000+, and it crashes as well

Olivier

Hi Olivier,

You must have an hash entry in the 'init string' box
for it to work properly e.g

hash32

New version out soon - Bugs fixed hopefully.

Please disregard my last few posts concening Winboard.
Leo's been testing with Winboard using Polyglot and I forgot
Apilchess is UCI only!

It all gets very confusing sometimes!

best,
Jim.