Re: [profiling] different results with different profiling methods
From: Tim Ward (tw2_at_ipaccess.com)
Date: 03/30/04
- Next message: Max: "Re: Javascript > Java Communication : not work on Mozilla Browser (jre1.5)"
- Previous message: Hans Kratz: "Re: JDK1.5, improved FOR loop and generics - How?!"
- In reply to: watzlaw wutz: "[profiling] different results with different profiling methods"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Tue, 30 Mar 2004 10:58:12 +0100
"watzlaw wutz" <watzlaw.wutz@gmx.net> wrote in message
news:c49im4$2fdfs1$1@ID-68449.news.uni-berlin.de...
>
> Could anyone explain me, why i get these results and which of them i
should trust. If you
> need more information or don't understand my english (sorry for it) let me
know.
Heisenberg's Uncertainty Principle - if you instrument code it will perform
differently. Two different methods of instrumentation will give two
different results. The more invasive a profiler is the more it will perturb
the results; the less invasive it is the less detailed the results.
Making sense of the output from profilers can be something of a black art -
years of experience help here, across multiple languages and platforms;
whilst profilers are certainly useful you should not rely on them definitely
always telling "the truth" (whatever that might be). Certainly after making
changes, and measuring the results with a profiler, you should measure the
results again with no instrumentation to see what improvements you have made
in real world conditions.
-- Tim Ward Brett Ward Limited - www.brettward.co.uk
- Next message: Max: "Re: Javascript > Java Communication : not work on Mozilla Browser (jre1.5)"
- Previous message: Hans Kratz: "Re: JDK1.5, improved FOR loop and generics - How?!"
- In reply to: watzlaw wutz: "[profiling] different results with different profiling methods"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
