Author: Aaron Gordon
Date: 02:50:15 09/27/01
Go up one level in this thread
On September 27, 2001 at 05:42:50, Gian-Carlo Pascutto wrote: >On September 27, 2001 at 05:33:06, Aaron Gordon wrote: > >>CFLAGS = -O2 -Qipo -G6 -Qunroll -Qprof_use -Qprof_dir c:\opt > >The newer Intel C's support -O3 as well, might want to try it. > >>After that it will create the optimized exe... I'm not a programmer or anything >>so I can't explain what the profiling does exactly, only guessing.. but it does >>seem to help a LOT. > >After profiling the compiler has a much better idea which branches will >be taken, what is worth inlining and what not, which data should be >grouped, etc... > >-- >GCP Ah ok.. Sounds good. Do you think profiling takes the L1/L2 cache into consideration? Such as a profiled exe on a Pentium 3 (fast 256K L2 cache)being run on a Celeron-2 (128k of slower L2) CPU may be slower than if it was profiled on the Celeron-2 itself. Do you think it's possible that could happen? Also what about the Katmai vs Coppermine P3's (or tbird vs duron), where the katmai has 512K 1/2 speed L2 vs coppermine's full speed 256K L2?
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.