Author: Vincent Diepeveen
Date: 12:40:55 05/10/02
Go up one level in this thread
On May 09, 2002 at 23:43:36, Dann Corbit wrote: >On May 09, 2002 at 23:22:08, Slater Wold wrote: > >>On May 09, 2002 at 23:11:22, Dann Corbit wrote: >> >>>On May 09, 2002 at 22:24:15, Vincent Diepeveen wrote: >>> >>>>On May 09, 2002 at 10:27:34, Frank Phillips wrote: >>>> >>>>>Intel Linux compiler version 6 is faster (about 5%) than Intel version 5 for me, >>>>>on a K7 1.4GHz (non XP) - Mandrake 8.2 with icc -O3 -ipo -prof_use >>>> >>>>and if you try on windows intel c++ 5.0.1 next: >>>> CFLAGS = -O3 -G6 -Qaxi -Qxi -Gr -Qprof_genx >>> >>>You left out this step: >>>Then run your program quite a while. I like perft, then an EPD test suite and >>>then some games. >>> >>>>then delete object files and try again: >>>> CFLAGS = -O3 -G6 -Qaxi -Qxi -Gr -Qprof_use >>> >>>I think K7 is OK for -G7 instruction mix flag, isn't it? >> >>No. -G7 is for P4's which have SSE2 among other things. > >But it won't emit those instructions. It just has an optimal instruction mix >for that chip (P4) which seems to work better for my Athlon. The -G flags are >really just for scheduling and for the instruction mix that works best on the >named chip. But it should run on any of the inferior chips. the G4 optimization creates crash of diepengine when i run the produced exe onto my K7-MP 1.2Ghz So i do not know whether the code is faster. All i know it is illegal code.
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.