Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Intel C++ Compiler

Author: Robert Hyatt

Date: 20:34:44 01/15/03

Go up one level in this thread


On January 15, 2003 at 21:39:59, David Rasmussen wrote:

>I'm a little bit unclear about the data created during phase one of the profiler
>guided optimization phase. Should I delete the dyn data whenever I make a
>change, or will it find out itself what data is still useful etc. ?
>
>Can I make more than one run to create profiler data?
>
>What I do now is this: Once in a while I do a profiler phase with an
>instrumented build, but most of the time, I just compile in feedback mode (phase
>3) even if I make a lot of changes etc.
>
>What is the correct way?
>
>/David


I am not sure about that either, so when I change something I get rid of
all the files.  however, you can make multiple runs and get multiple .dyn
files and it will "combine" 'em when optimizing to get better branch
probabilities...

I have a "make profile" in my makefile that clears the files, compiles
with the prof_gen option, runs a set of positions, touches all the source
files and compiles again with prof_use...




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.