RE: [Bug-gnubg] 50% decrease in evaluation speed

[Ian Shaw]

I remember that when Thomas Hauk used gnubg in his thesis on search trees with 
chance nodes, he converted the floating point operations to integers. The 
stated purpose was to avoid any inconsistencies due to rounding errors, but he 
also claimed that the integer maths had a slightly better performance.

Is there any benefit in looking into this? I believe that his version of gnubg 
was pre-SSE, so that work may have surpassed any benefit from integer maths.

-- Ian