Author: Jürgen Hartmann
Date: 06:43:21 02/17/99
Go up one level in this thread
Bravo, I find that brilliant. An ideal protection. Force people to understand what you are doing. Jürgen >changed in the wrong places, the SMP search breaks in subtle ways. It causes >me a lot of grief as when I make changes and don't check these 'tripwires' I >trigger them myself and do a lot of debugging. > > >I've done one thing already that I suppose won't hurt to reveal. Way back in >the SMP development, I thought about this problem. And I made a couple of >'assumptions' that are wrong about some things I do in the SMP search. These >'wrong' things can be avoided by subtle programming things, but if the code is >I won't say more, other than that at least 'copying' the SMP code line for line >is guaranteed to produce two nice 'glitches' that will produce amusement when >they show up. But they are subtle. I only wish it was 'easy' to make the >'copy' act up upon demand... but the bugs are more subtle than that, >unfortunately. > >If someone studies the SMP code _very_ carefully, the problems can be exposed. >But that is the point... because at least they have to understand the algorithm >to make it usable, or else run the risk of some 'interesting behavior'... So I >have taken one step at least, but only WRT the SMP stuff. (one of these >problems was simply a bug that I had to find before the current SMP algorithm >was functional in early 15.x versions. But it was cute enough that I decided >that this was the 'perfect' protective mechanism, and I just chose to fix the >'symptom' rather than the actual problem.) > >Let's just see what happens...
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.