Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Critter 1.6 x64 engine crash
- - By Peter Grayson (****) Date 2012-06-21 09:16
Critter 1.6 x64 engine crashed during a 60 game Noomen2012 match against Stockfish 2.2.2.

dr.watson reported a divide by zero error when Critter started a new game at game 26. Engine failed to start.

Application exception occurred:
        App: D:\Chess Engines\critter_1.6_win\Critter_1.6_64bit.exe (pid=1116)
        When: 21/06/2012 @ 07:31:40.562
        Exception number: c0000094 (divide by zero)
followed by a load of x86 instructions indicating function etc. where error occurred.

Game in Windows XP 64 bit with Large Pages enabled.
Q9550 CPU O/C'd to 3.53 Ghz. 2Gb Hash. Chessbase GUI 11. Ponder on.
Match run in Autoplayer using two PC's cross connected using Serial Port Com1.

Anyone else had a problem?

PeterG
Attachment: Critter1_6vsStockfish222.cbv - Games (40k)
Parent - - By Uly (Gold) Date 2012-06-21 10:18
Do you have the game that was running when it happened? It would be useful to turn on Critter's logging on and reproduce it.

(Critter has been improving steadily but I don't expect it to support division by zero any time soon :lol:)
Parent - - By Peter Grayson (****) Date 2012-06-21 12:11 Edited 2012-06-21 12:20
Critter completed game No.25. The GUI was running as the master, sent out the database moves for game 26 to the slave GUI but Critter engine failed to start, confirmed by Task Manager. GUI then timed out and Windows error message appeared saying Critter had encountered a problem. Been in touch with some other enthusiasts and they have experienced similar problem when running test sets or games. May be Chessbase GUI related but thought R.V. had sorted those issues out in 1.4 update release.

As for the crash, it was about the engine not starting rather than failing during a game. It had won the previous game with mate. I've attached the game file in the original post and embedded the first failure window.

PeterG
Parent - - By Banned for Life (Gold) Date 2012-06-21 12:45
Stockfish was using the rope-a-dope technique. Even though Critter 1.6 was ahead on the scorecard, it couldn't come out for the 26th round, so it lost the match... :lol:
Parent - - By Peter Grayson (****) Date 2012-06-21 13:22
Yes ... very good!:lol:

On a more serious note, playing on 2 PC's should prevent any interference between engines although I seem to recall in the past that Ed Schroeder dropped the Autoplayer facility from Rebel claiming other programmers were using it to read Rebel's opening book so I guess it may be possible for foul play through that route. First time for some while though that this has happened in my engine matches, despite being run in the quirky Chessbase GUI.  Apologies to Ed if it was another program.

PeterG
Parent - - By Banned for Life (Gold) Date 2012-06-21 13:33
By coincidence, I've been playing 1'1" blitz matches on a single quad between SF2.2.2JA and Critter1.6 on the F12 gui almost continuously since Richard's release, and haven't had any problems...

Of course if you are playing a long match using autoplayer, and don't mind losing a lot of short games, you could do what Ed suggested. Eros was probably on the other end of the line! :lol:
Parent - - By Uly (Gold) Date 2012-06-21 15:24

> By coincidence, I've been playing 1'1" blitz matches on a single quad between SF2.2.2JA and Critter1.6 on the F12 gui almost continuously since Richard's release, and haven't had any problems...


Very nice! Are you planning to repeat the match this time turning Critter Session File on? It would be cool to know if its performance would increase with "Learning" on (RomiChess had some "Monkey-See Monkey-Do" learning implementation that would bring 200 elo, but it was a weaker engine learning from stronger opponents and winning time on the clock by playing instantly.)
Parent - - By Banned for Life (Gold) Date 2012-06-21 15:31
Learning has been on the entire time. Sometimes it helps, but sometimes Critter learns stupid things! :lol: How did that line go from Forest Gump? Stupid is as stupid does...
Parent - By InspectorGadget (*****) Date 2012-06-21 17:34

> but sometimes Critter learns stupid things! :lol:


:lol:
Parent - - By Uly (Gold) Date 2012-06-22 06:11
Very nice! Are you planning to repeat the match this time turning Critter Session File off? It would be cool to know if its performance would decrease with "Learning" off.

:lol:
Parent - - By Banned for Life (Gold) Date 2012-06-23 06:43
I am using it for an endgame study...
Parent - By Uly (Gold) Date 2012-06-23 08:38
Very nice! I'm glad finally someone is playing games without being interested in the elo performance one program has over another! :yell:
Parent - By Razor (****) Date 2012-06-21 16:57
Agreed - testing on two PCs enables a more 'real-life' test environment to develop a testing framework on.
Parent - - By Richard Vida (**) Date 2012-06-21 15:08
In the whole code there is only one division with unchecked divisor (in time management code). It is possible to reproduce the crash by specifying some unusual TC values, but I doubt any GUI would do that:

ucinewgame
go wtime 1 btime 1 movestogo 1
Parent - By Peter Grayson (****) Date 2012-06-21 16:01 Edited 2012-06-21 17:44
From what I recall of tests carried out with the Houdini 1.5 engine it was found that the Chessbase GUI does not issue a "New Game" instruction at the start of Autoplayer Games, only a "Clear hash" instruction.  Below is an extract from the ucilogfile that showed this up when starting a 40 moves in 5 mins repeating time control multi-game match using an opening database.

Have run 1.4 version without issue for several hundred games under same conditions.

Note the clear hash instruction appears after the first move and is probably the cause of the first to move engine ponder stopping immediately after it makes the move. Ponder is not affected subsequently. Chessbase seems to use its own convention for some of this stuff.

PeterG

[2011/01/09 03:44:57] position startpos moves g1f3 g8f6 g2g3 g7g6 f1g2 f8g7 e1g1 e8g8 d2d3 d7d6 e2e4
[2011/01/09 03:44:57] go movestogo 35 btime 300000 wtime 300000
[2011/01/09 03:45:05] > bestmove c7c5 ponder e4e5
[2011/01/09 03:45:05] setoption name Clear Hash
[2011/01/09 03:45:37] position startpos moves g1f3 g8f6 g2g3 g7g6 f1g2 f8g7 e1g1 e8g8 d2d3 d7d6 e2e4 c7c5 b1c3
[2011/01/09 03:45:37] go movestogo 34 btime 291000 wtime 300000
[2011/01/09 03:45:47] > bestmove b8c6 ponder a2a3
[2011/01/09 03:45:47] position startpos moves g1f3 g8f6 g2g3 g7g6 f1g2 f8g7 e1g1 e8g8 d2d3 d7d6 e2e4 c7c5 b1c3 b8c6 a2a3

Edit: I thought this had a familiar ring to it ...
http://open-chess.org/viewtopic.php?f=7&t=873
Up Topic The Rybka Lounge / Computer Chess / Critter 1.6 x64 engine crash

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill