Author: Robert Hyatt
Date: 11:59:04 12/10/03
Go up one level in this thread
On December 10, 2003 at 01:48:36, Russell Reagan wrote: >On December 10, 2003 at 01:22:23, Slater Wold wrote: > >>Ususally you'll see 'like results'. > >By "like results" do you mean that a program that doesn't benefit from the >64-bit registers (a non-bitboard program would fall into this category) will >perform similar to its 32-bit counterpart? > >In other words, we shouldn't expect these kinds of speedups for 0x88-style >programs? Extrapolation is impossible. There are three issues that speed up programs: (1) the processor is a bit more efficient in several places. This helps everybody. (2) the processor has 2x as many registers, which also helps everybody since you don't run into "register jams" as frequently or as painfully. (3) the processor works on 64 bits rather than 32 bits. This only helps those programs that use 64 bit values. At the moment everybody is doing this to some extent with hashing, but for bitboard programs, most of the internal logic is depending on 64 bits as well, so they should/will get a bigger gain relative to programs that still use 32 bits. Trying to characterize which of the above 3 is the most important is _really_ hard.
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.