Author: Vincent Diepeveen
Date: 05:07:10 09/12/00
Go up one level in this thread
On September 11, 2000 at 18:36:14, Dieter Buerssner wrote: >On September 11, 2000 at 18:22:13, Vincent Diepeveen wrote: > >>I compiled latest snapshot from the gcc compiler again, >>and measured speed difference using -O9 -fexpensive-optimizations >>-mcpu=pentiumpro -march=pentiumpro etcetera > >Have you tried gcc with -O or with -O2. All my programs, which I have checked >for execution speed (chess program and numerical/quantum chemical simulations), >run *slower*, when compiled with gcc and -O3 on x86. (I think -O9 is the same as >-O3). Especially hurting seems to be the automatic inlining of gcc (that is not >enabled by -O2). This can also be seen by looking at the assembler output. >However, manual inlining of critical and fast functions by the inline keyword >can help. Matters are quite different, when using a CPU with many registers (I >think almost all contemporary CPUs other than x86 based). For DIEP it doesn't matter whether i use -O2, -O3 or -O9 it makes shit difference already for the past so many years. However using PGCC with -O3 or above slows down diep 15% at least. The register efficiency or something from gcc hasn't improved past 5 years. >- Dieter
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.