This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Q: Profile-directed optimizations
- To: law at cygnus dot com, ahuang at ibmoto dot com
- Subject: Re: Q: Profile-directed optimizations
- From: N8TM at aol dot com
- Date: Wed, 2 Sep 1998 20:23:58 EDT
- Cc: egcs at cygnus dot com
In a message dated 9/2/98 7:42:16 AM Pacific Daylight Time, law@cygnus.com
writes:
> > 1) it is difficult to do,
> > 2) it wouldn't buy much performance, or
> > 3) just that no one has gotten around to it?
> #1, #3.
>
> Some folks might also claim #2. I'm not convinced either way on whether
> or not profile directed feedback helps real code.
I imagine #1 is particularly true if it's never been done for a multiple
architecture compiler, and that would contribute to #2. I've tried profile-
directed optimization on both real codes and Livermore Kernels, and what was
gained in one section was lost in another.