PolyGlot getting release ready ...

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 07 April 2004 12:42:27:


I am planning for a 1.0 release next week. I can make changes next
week-end but bear in mind that Friday is a bank holiday in England (as
well as Monday) ... In other words, I need to know about problems
tomorrow at the latest.
I think the executable Dann made available is in "release" mode, which
means it should be complaining much less about bugs in the engines.
It is hopefully faster as well.
Leaving aside the "unloading" and other known problems for now, I
would like to have comments about performance.
As I explained before, performance is not an issue at all for Linux
and other POSIX operating systems. However on Windows CYGWIN.DLL has
to do some multitasking and I/O work, which means it can slow down
PolyGlot noticeably. Can anybody give me some figures about CPU time
consumed by PolyGlot during a game? Does it seem acceptable?
Info about compatibility with other interfaces than Winboard is also
welcome. If you have time to do so, please try PolyGlot with
xboard-compatible interfaces such as Arena (yes I know Arena has
native support for UCI), SCID, ChessMaster (supports xboard?), etc ...
Also any other xboard-compatible utilities you might find useful such
as EPD-suite testing, connection with Internet servers and so on ...
Do that only if you have some use for it.
I expect most of these "alternative" solutions *not* to support
version 2 of the xboard protocol, but PolyGlot has partial support for
version 1.
Fabien.

[Volker Pittlik]

Geschrieben von:/Posted by: Volker Pittlik at 07 April 2004 14:42:19:
Als Antwort auf:/In reply to: PolyGlot getting release ready ... geschrieben von:/posted by: Fabien Letouzey at 07 April 2004 12:42:27:

...

Info about compatibility with other interfaces than Winboard is also
welcome. If you have time to do so, please try PolyGlot with
xboard-compatible interfaces such as Arena (yes I know Arena has
native support for UCI), SCID, ChessMaster (supports xboard?), etc ...

Scid:
Working well: Shredder, SOS-4
Not working: Alfil
Crashing: Piranha
Volker

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 07 April 2004 15:27:13:
Als Antwort auf:/In reply to: PolyGlot getting release ready ... geschrieben von:/posted by: Fabien Letouzey at 07 April 2004 12:42:27:

My report on UCI-only engines running under Polyglot.

The full list of UCI-only engines:
----------------------------------
Piranha 0.5 no problem
T-Rex 1.8.5 no problem
SOS 4 for Arena no problem
Fruit 1.0 no problem
List_512 no problem
ALFIL 403.1 no problem (with "Send Done = false")
Silke Chess 1.2.1209 no problem but the engine GUI stays in RAM after the match.
DelphiMax 2.8 no problem but the engine stays in RAM after the match.
FUSCh 1.09 it cannot run even though 1.07 and 1.08 can run with
/firstProtocolVersion=1 and
/secondProtocolVersion=1. Also both versions cause
an error message on match completion.
Little Goliath Nemesis it runs fine with Use_Position_Learning = false. But
when learing is on and after it plays a game and
restarts - then it exits. After I deleted the file
NEMESIS.LRN and switched off learing it can run
again.

Each engine played 4-game match at time control 20".
Igor

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 07 April 2004 15:41:40:
Als Antwort auf:/In reply to: report on UCI-only engines geschrieben von:/posted by: Igor Gorelikov at 07 April 2004 15:27:13:

My report on UCI-only engines running under Polyglot.

The full list of UCI-only engines:
----------------------------------
...
Silke Chess 1.2.1209 no problem but the engine GUI stays in RAM after the match.
DelphiMax 2.8 no problem but the engine stays in RAM after the match.
Little Goliath Nemesis it runs fine with Use_Position_Learning = false. But
when learing is on and after it plays a game and
restarts - then it exits. After I deleted the file
NEMESIS.LRN and switched off learing it can run
again.

Each engine played 4-game match at time control 20".

I am willing to help about the unloading problem, but nobody reports log files so I have no idea what to look for.
But I guess that Silke Chess and DelphiMax are not written in C, right?
Maybe there are things I should know about "alternate" languages (concerning I/O) ...
I would like to have detailed log files for Little Goliath (that's Winboard log + PolyGlot log + Goliath log if possible).
Do you mean 20 secs / game? wow!
Especially impressive because UCI engines are very verbose during the search.
I am actually thinking about an option to switch off info parsing, to see if it helps performance.
Fabien.

[Günther Simon]

Geschrieben von:/Posted by: Günther Simon at 07 April 2004 15:58:41:
Als Antwort auf:/In reply to: report on UCI-only engines geschrieben von:/posted by: Igor Gorelikov at 07 April 2004 15:27:13:

My report on UCI-only engines running under Polyglot.

The full list of UCI-only engines:
----------------------------------
Piranha 0.5 no problem
T-Rex 1.8.5 no problem
SOS 4 for Arena no problem
Fruit 1.0 no problem
List_512 no problem
ALFIL 403.1 no problem (with "Send Done = false")
Silke Chess 1.2.1209 no problem but the engine GUI stays in RAM after the match.
DelphiMax 2.8 no problem but the engine stays in RAM after the match.

I guess you exchanged Silke and DelphiMax accidentally,
because Silke has no GUI but Delphi has?
Best regards,
Günther

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 07 April 2004 16:06:03:
Als Antwort auf:/In reply to: Re: report on UCI-only engines geschrieben von:/posted by: Fabien Letouzey at 07 April 2004 15:41:40:

I would like to have detailed log files for Little Goliath (that's Winboard log + PolyGlot log + Goliath log if possible).

Yes, it is possible and if you need them I send them to you.
But I'm sure it is the LG problem and only author can fix it. The same
behavour is under Arena.

Here are the tails of winboard and Polyglot logs:
================================================
Winboard's
==========
2030 (second: feature usermove=1
2030 )second: accepted usermove
2030 (second: feature variants="normal"
2030 )second: accepted variants
Fatal Error: Error: second chess program (polyglot LGNemesis.ini) exited unexpectedly
GameEnds(0, (null), 2)
17960 )first : quit
17960 )second: quit
Polyglot's
==========
( XBOARD accepted time
( XBOARD accepted usermove
( XBOARD accepted variants
( ENGINE EOF
And here is the Windows standard error message:
-----------------------------------------------
This program performed an illegal operation and will be shut down.
LGNEMESIS caused an invalid page fault in
module LGNEMESIS.EXE at 0167:00418422.
Registers:
EAX=00178824 CS=0167 EIP=00418422 EFLGS=00010206
EBX=00000004 SS=016f ESP=0068ad20 EBP=00000383
ECX=052349c9 DS=016f ESI=00000000 FS=46ff
EDX=00000000 ES=016f EDI=004393ad GS=0000
Bytes at CS:EIP:
89 08 8b 15 18 f4 44 00 8a 44 24 0b 88 42 09 8b
Stack dump:
052349c9 f961f603 0701fff3 089b0001 00402333 00000000 00000001 007b0b70 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
Igor

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 07 April 2004 16:08:27:
Als Antwort auf:/In reply to: Re: report on UCI-only engines geschrieben von:/posted by: Günther Simon at 07 April 2004 15:58:41:

I guess you exchanged Silke and DelphiMax accidentally,
because Silke has no GUI but Delphi has?
Best regards,
Günther

You are right: the cut and paste plays this thing. Thank for correction, Günther.
Igor

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 07 April 2004 16:11:20:
Als Antwort auf:/In reply to: Re: report on UCI-only engines geschrieben von:/posted by: Igor Gorelikov at 07 April 2004 16:06:03:

I would like to have detailed log files for Little Goliath (that's Winboard log + PolyGlot log + Goliath log if possible).

Yes, it is possible and if you need them I send them to you.
But I'm sure it is the LG problem and only author can fix it. The same
behavour is under Arena.
This program performed an illegal operation and will be shut down.

I see ...
Nothing I can do about that
Fabien.

[Norm Pollock]

Geschrieben von:/Posted by: Norm Pollock at 07 April 2004 17:02:23:
Als Antwort auf:/In reply to: report on UCI-only engines geschrieben von:/posted by: Igor Gorelikov at 07 April 2004 15:27:13:

I'm using jori's tournament manager. I have trouble with polyglot and uci engines sos4 and list-5.12. The trouble is not likely from polyglot because fruit with polyglot/bookthinker works very well.
With sos4/polyglot the engine plays very well but sometimes refuses to leave memory after a game is finished.
With list-5.12 and polyglot/bookthinker, the engine plays very well but does not report a result to the tournament manager. It's games end with "*".

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 07 April 2004 17:12:09:
Als Antwort auf:/In reply to: Re: report on UCI-only engines geschrieben von:/posted by: Norm Pollock at 07 April 2004 17:02:23:

I'm using jori's tournament manager. I have trouble with polyglot and uci engines sos4 and list-5.12. The trouble is not likely from polyglot because fruit with polyglot/bookthinker works very well.
With sos4/polyglot the engine plays very well but sometimes refuses to leave memory after a game is finished.
With list-5.12 and polyglot/bookthinker, the engine plays very well but does not report a result to the tournament manager. It's games end with "*".

I need log files to know how this can happen.
Interface logfile (is it Winboard?) + PolyGlot log file, SOS log files can't hurt.
This is a known bug in List that occurs mostly near the end of the game.
Fabien.

[Norm Pollock]

Geschrieben von:/Posted by: Norm Pollock at 07 April 2004 18:38:53:
Als Antwort auf:/In reply to: Re: report on UCI-only engines geschrieben von:/posted by: Fabien Letouzey at 07 April 2004 17:12:09:

I'm using jori's tournament manager. I have trouble with polyglot and uci engines sos4 and list-5.12. The trouble is not likely from polyglot because fruit with polyglot/bookthinker works very well.
With sos4/polyglot the engine plays very well but sometimes refuses to leave memory after a game is finished.
With list-5.12 and polyglot/bookthinker, the engine plays very well but does not report a result to the tournament manager. It's games end with "*".

I need log files to know how this can happen.
Interface logfile (is it Winboard?) + PolyGlot log file, SOS log files can't hurt.
This is a known bug in List that occurs mostly near the end of the game.
Fabien.

I rechecked and sos4 is now working fine. thanx.