PB 4.0 Performance
Posted: Sun Feb 05, 2006 2:44 pm
Because PB 4.0 goes for thread safety and new types, I was a bit afraid that all this new stuff would hurt performance somehow. But now that my chess program has been converted to the new PB 4.0 syntax, I have to admit that I'm quite impressed. Although some critical parts of the game are coded in assembler, the new compiler gives a 15% speed improvement. The size of the executable grows by 2%, but that's near to nothing. But there's a problem: one test position gives a different result. There must be a difference between 3.94 and 4.0 in some calculation or comparison. To find such a thing is a coder's nightmare... 