This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: performance regressions, profiling _of_ gcc
- From: Michael Matz <matz at suse dot de>
- To: <jan at etpmod dot phys dot tue dot nl>
- Cc: <gcc at gcc dot gnu dot org>
- Date: Wed, 9 Apr 2003 18:18:00 +0200 (CEST)
- Subject: Re: performance regressions, profiling _of_ gcc
Hi,
On Thu, 10 Apr 2003, Jan Van Dijk wrote:
> Thank you (and Steven) for your quick replies. That message contained
> the info I needed, I am sorry I missed it in my search.
>
> One question remains: I was a bit surprised that the desire to enable
> profiling cannot be expressed via the configure command (--enable-profile,
> say).
This might be usefull, but nobody did it yet ;)
> Configure could pass the -pg flag to stage 2 and 3 when bootstrapping,
> it seems.
You probably don't want to do that. There is not much sense in running a
full bootstrap with a profiling compiler. This is slightly different than
if one wants to do that for creating a compiler which used profile
feedback (like Honza's make profiledstrap or how it was called).
But for purely development work you don't bootstrap, but instead simply
make a cc1(plus) which is profilable, i.e. no multiple stages involved.
Ciao,
Michael.