Author: Ricardo Gibert
Date: 14:37:24 04/13/02
Go up one level in this thread
On April 13, 2002 at 16:22:08, Richard Heldmann wrote: >How important is it which compiler is used to compile a chess program? Does it >create a significant ELO improvement/loss? I believe someone has reported a 30% speedup by compiling with MSVC6 instead of GCC. This is worth about 15 rating points. If 15 rating points is significant to you then which compiler you use is important. The 15 rating point figure is computed by assuming every doubling of speed is worth about 40 rating points. This is a common type of rule of thumb. Some whould say 50, in which case you would expect a 19 rating point improvement. The formula is: Delta R = c*ln(s), where c = d/ln(2), s = speedup factor, d = rating differential for each doubling of speed. For example, for the 15 rating point figure, I used d = 40, which gives c = 58 = 40/ln(2). Now with s = 1.3 for a 30% speedup, you have 15 = 58*ln(1.3). If d = 50, then c = 72. Of course, all the above calculations are approximate and are only intended to give a rough idea.
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.