Terminating the ChessGUI project.

[Martin Thoresen]

I have not found the reason why first 0.198 crashed.

Matthias.

Well, as you know, the test tournament has been running flawlessly so far - not one single crash.

I can't explain why, either. I just hope the crash will come soon so that we can debug it.

Best Regards,
Martin

[Matthias Gemuh]

I have not found the reason why first 0.198 crashed.

Matthias.

Well, as you know, the test tournament has been running flawlessly so far - not one single crash.

I can't explain why, either. I just hope the crash will come soon so that we can debug it.

Best Regards,
Martin

Sometimes it not good if a progam does not crash.

Matthias.

[mhurd]

Here is the version that should be used to test for crashes and time forfeits:
http://www.hugedrive.com/published/WG/s ... =-3eb0543f
It is an improved version of the one that crashed.
BTW, when testing at fast TC, "Time Left" is more important in pgn than "Move Time".

I am running a 396 game tournament with the above version, nearly 200 games so far, no crashes and seems to be running ok.

Mike

[mhurd]

The 396 games have now finished. Time control was game in 60 seconds + 1 second.

Results :-

121.0 Houdini 1.03a x64 4_CPU
111.0 Deep Rybka 4 x64
97.5 Stockfish 1.8 JA 64bit
66.5 Critter 0.80 64-bit

The program appeared to run fine and there were no crashes.

Win 7 64 bit
4 Gb Ram


Mike

[Jouni]

In version 199 Stockfish still loses most games by time (60/60 level ) ...

Jouni

[Matthias Gemuh]

In version 199 Stockfish still loses most games by time (60/60 level ) ...

Jouni

Default timing is strict.
Can you try relaxed timing and tell us the outcome ?

[hgm]

Is there a UCI engine that sends the time stamp of its move as an info string, so that it ends up in the GUI log (e.g. winboard.debug) together with the move time stamped at the receiving end, so that we could determine the exact communication delay?

This seems a much more informative (and quicker) method for determining where the problem lies than playing games, and counting the number of time losses...

[Jouni]

Done: no more time losses! I haven't followed all discussion, so what's
"relaxed timing" meaning?

Jouni

[hgm]

That the GUI is forgiving when you only exceed the time by a little amount, and does not flag you. But when playing on-line, the ICS calls the shots, and it will be strict, so you would still lose there...

[Matthias Gemuh]

That the GUI is forgiving when you only exceed the time by a little amount, and does not flag you. But when playing on-line, the ICS calls the shots, and it will be strict, so you would still lose there...

The so-called relaxed timing in ChessGUI works as follows:

when ChessGUI reads in-coming data, it stuffs it into a FIFO buffer.
If a move is in that data, ChessGUI notes the time (PipeTime in debug files), then it uses PostMessage() to alert the routine that will process that data.
The time that PostMessage() takes is measured as lag (see debug files).
On my WinXP32, this lag is zero ms most of the time, but sometimes 16 ms.
In relaxed timing, ChessGUI assumes that the same amount of lag occurs when GUI sends data to engine, so it adds that lag to engine's time.
This lag is basically a measure of how much Windows is ignoring the GUI and engines.

Relaxed timing is of course not for engine authors.

Matthias.