This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: 3.0 high priority: optimization/3083: C++ frontend consumesinacceptableamounts of CPU with -O3
- To: Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>, gcc at gcc dot gnu dot org
- Subject: Re: 3.0 high priority: optimization/3083: C++ frontend consumesinacceptableamounts of CPU with -O3
- From: Paolo Carlini <pcarlini at unitus dot it>
- Date: Sat, 09 Jun 2001 18:53:29 +0200
- Organization: Universita' della Tuscia
- References: <Pine.BSF.4.33.0106091208110.82995-100000@deneb.dbai.tuwien.ac.at>
- Reply-To: pcarlini at unitus dot it
Hi,
Gerald Pfeifer wrote:
> On Fri, 8 Jun 2001, Paolo Carlini wrote:
> > please excuse me in advance for the naivety of my observation: are you
> > fully aware of the fact that in the time span between Mitchell's fixes
> > of about two monts ago and the following patch (of about one month ago):
> >
> > http://gcc.gnu.org/ml/gcc-patches/2001-05/msg00233.html
> >
> > the -O3 option was broken?? (i.e. -O3 compilations were faster than
> > they really would have been if -O3 was really working...)
>
> Thanks for the pointer (and no, I wasn't aware of that). It seems,
> however, that the performance regression I am seeing is unrelated.
Ok, and sorry for having diverted you from the real problem which indeed
appeared much later...
Well, another couple of my naive thoughts: there is the coincidence that the
important eh merge happened exactly between the 12th and 13th of May... (of
course I ignore completely if processing the eh information is heavy or not
for the compiler...) I wonder if your last successful compilation is before or
after that...
Also, the 14th of May the important libstdc++ merge (again, I ignore
completely if your generate.C uses much of the library neither I timed if the
latest libstdc++ developments slowed down the compilation...)
Good luck and sorry again...
P.