Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: quarkx v monsoon-ccct4

Author: David Rasmussen

Date: 04:03:23 01/21/02

Go up one level in this thread


On January 20, 2002 at 22:16:08, Scott Gasch wrote:

>
>I'm more than open to suggestions about what else to try -- I'd be grateful for
>any ideas.
>

I haven't actually implemented this myself yet, but I will:

I think it would be a good idea if the engine made a compact log different
things (well, most do already), but specifically off how many nodes was
searched/pondered. Because then you can _always_ recreate the exact same
environment for your engine, as when the bug happened. You can even have it done
automatically, by having a command that let's the engine process the log, that
is, reads how many nodes was searched now, and searches exactly that many nodes.

Another thing I have done is a command to save and load the hashtable to a file.
I was able to reproduce a bug a couple of weeks ago, even though the environment
wasn't exactly the same. I just searched for 5 seconds on every move, until the
move where the bug happened. Luckily, it made the same mistake, the bug was in
effect. But this was at move 80 or so, and it would be a pain to recreate this
environment everytime I started the debugger. So I saved the hashtable, and
everytime I started the debugger, I loaded the hashtable also. Of course this
can be done more systematically by not just saving the hashtable, but the entire
state of the program. Pawn hashtables, repetition information etc. I eventually
tracked down the bug in about 12 hours or so.

/David



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.