This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: some profiling numbers
- From: Joe Buck <jbuck at synopsys dot com>
- To: graydon hoare <graydon at redhat dot com>
- Cc: gcc at gcc dot gnu dot org
- Date: Wed, 25 Jun 2003 12:06:41 -0700
- Subject: Re: some profiling numbers
- References: <3EF9DE6D.3030605@redhat.com>
On Wed, Jun 25, 2003 at 01:39:57PM -0400, graydon hoare wrote:
> I've recently done some profiling and analysis of gcc and g++, using the
> oprofile hardware-assisted profiling module in linux 2.5. I collected
> stats on a variety of versions of gcc (some only of historical interest)
> and some different branches and workloads, and characterized some
> aspects of the cache and branch-mispredict penalties in addition to
> cycle hotspots.
>
> the results of the work are posted at
> http://people.redhat.com/graydon/gcc-optimizing/
> and http://people.redhat.com/graydon/g++-report/
Beautiful work. I am encouraged to see that gcc-3.3 builds the Linux
kernel faster than gcc-3.2.x, even if we are still slower than 2.95.x.
At least *something* is going in the right direction.
It would be nice to see graphs for C++ compilation of some large program
analogous to the graphs for the Linux kernel. Unfortunately, there is a
huge "gotcha" in that libstdc++ is completely different for 3.x vs 2.x,
so we would be comparing different code. This means it would require more
thought to figure out how to do the benchmark properly. On the other
hand, we could ignore the issue and just compare releases from 3.0.x on.