Author: Vincent Diepeveen
Date: 14:28:44 02/16/05
Go up one level in this thread
On February 16, 2005 at 15:51:03, Eugene Nalimov wrote: >On February 16, 2005 at 15:26:50, Robert Hyatt wrote: > >>On February 16, 2005 at 15:15:53, Frank Phillips wrote: >> >>>On February 16, 2005 at 13:19:46, Robert Hyatt wrote: >>> >>>>On February 16, 2005 at 13:09:14, Frank Phillips wrote: >>>> >>>>>Has anybody got any experience with g++ 3.4 for amd64 (x86_64) - for Linux? >>>>> >>>>>I have been using the profile generated optimisation option, but the code it >>>>>produces is no faster then with simple -O3. >>>>> >>>>>I simply compile with -fprofile-generate >>>>>then run, >>>>>then recompile with -fprofile-use. >>>>> >>>>>The relevant *.gcno, *.gcda files are produced. Must be doing something >>>>>wrong..... >>>>> >>>>>Frank >>>> >>>> >>>>It has never worked well for me. In fact, it usually crashes when I try to do >>>>this for crafty. ICC (Intel's compiler) works fine and that is what I use >>>>myself. >>>> >>> >>>Oh... the last version of ICC I used produced fast 32 bit code. Does it now do >>>amd 64 bit ? >>> >>>Frank >>> >> >>No. Sorry. I completely overlooked the AMD-64 in your text as that has become >>so common to see over the past year. ICC is 32 bits only although it will be 64 >>bit before long... > >ICC 8.1 supports x86, x64, and IPF. Does it assume 1 port for L1 cache reads or 2? AMD x86-64 has 2 @ 3 cycles. intel has 1 @ 4 cycles (Prescott). branch prediction works different too. pretty relevant optimizations for AMD. >Thanks, >Eugene > >> >>> >>>>You need a good test set of data so that it can accurately predict each branch >>>>and move code to optimize branch prediction at run-time... The positions you >>>>test need to pretty accurately represent what happens in a real game.
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.