Author: Richard Pijl
Date: 05:19:55 07/12/02
Go up one level in this thread
On July 12, 2002 at 08:04:55, Christophe Drieu wrote: >Strange compiler results: > >I tried Visual C++ 6 and DigiatlMars compiler with my TRex engine. > >I got a position and try the 2 engine with 2 algorithm with depth=7 : > >VC++ (optimization maximum speed): > Trex alphabeta + null move > move b6a5 > -0.41 > 3.149.862 positions in 28 seconds > > alphabeta > move d5d4 > +0.20 > 15.540.730 positions in 2 minutes and 31 seconds > 102.918 Nd/s > > >sc -0: alphabeta + null move > move b6a5 > -0.41 > 2.606.646 positions in 24 seconds > > alphabeta > move d5d4 > +0.20 > 48.662.903 positions in 7 minutes and 41 seconds > 152.559 Nd/s > >What do you think about that ? It says that you are probably using non-portable pieces of code, have a bug (e.g. referencing uninitialized memory) or that you give the compilers different options (e.g. data alignment differences could get you a different hashtable entry size so you can stack more in the same amount of memory. As described some threads below search is quite resilient for errors in eval so that might explain that both programs yield the same move and value. Richard.
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.