Author: Gerd Isenberg
Date: 06:14:25 11/27/03
Go up one level in this thread
On November 27, 2003 at 07:55:17, Gian-Carlo Pascutto wrote: >On November 27, 2003 at 06:05:04, Daniel Clausen wrote: > >>So using unsigned 32-bit would be a good idea, right? > >Apparently not. I tried replacing int with unsigned int in >Deep Sjeng's movegen and it got slower (using MSVC for AMD64)! > >-- >GCP hmm, interesting. That contradicts the statement from amd's optimization guide. Maybe ms did something in their code generator to make unsigned even slower ;-) I would expect something like this: signed int int idx = ...; int elem = someArray[idx]; movsx rax, [idx] ; sign extension 32->64 bit mov edx, [someArray + 4*rax] ; array access versus unsigned int unsigned int idx = ...; int elem = someArray[idx]; mov eax, [idx] mov edx, [someArray + 4*eax] ; implizite zero extension Cheers, Gerd
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.