Re: optimization using profiling

"John Williams" <jrw@pobox.com>
15 Nov 1998 13:29:38 -0500

          From comp.compilers

Related articles
optimization using profiling iyossi@my-dejanews.com (1998-11-12)
Re: optimization using profiling mikey@ontek.com (1998-11-15)
Re: optimization using profiling jrw@pobox.com (John Williams) (1998-11-15)
Re: optimization using profiling joachim.durchholz@munich.netsurf.de (Joachim Durchholz) (1998-11-15)
| List of all articles for this month |
From: "John Williams" <jrw@pobox.com>
Newsgroups: comp.compilers
Date: 15 Nov 1998 13:29:38 -0500
Organization: The University of Texas at Austin, Austin, Texas
References: 98-11-078
Keywords: optimize

<iyossi@my-dejanews.com> wrote in message 98-11-078@comp.compilers...
[snip]
>How? What can be done after profiling that can't be done without it ?
>To get a real improvement, I guess I must run my application to cover
>ALL the possibilities (all branches in code), or I'll get worse
>performance, since the compiler will think that some code is never or
>almost never being executed, right?


No; if some branches are rarely taken, this is exactly what profiling is
trying to determine.


Some very important optimizations (e.g. procedure inlining) depend on
finding the most common branches. Inlining commonly-called procedures can
dramatically improve performance, but inlining the wrong procedures only
causes code bloat. Profiling can be especially useful for object oriented
languages where viritual procedures make inlining much more difficult to do
effectivlely. The UW Cecil/Vortex project is a good example of using
profiling to guide optimizations like this.


jw


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.