This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: x86_64, 4.4 regression on optimization
- From: "Richard Guenther" <richard dot guenther at gmail dot com>
- To: "Daniel Franke" <franke dot daniel at gmail dot com>
- Cc: fortran at gcc dot gnu dot org
- Date: Mon, 30 Jun 2008 17:49:59 +0200
- Subject: Re: x86_64, 4.4 regression on optimization
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:date:from:to :subject:cc:in-reply-to:mime-version:content-type :content-transfer-encoding:content-disposition:references; bh=6KgMhOAP7r7nICLk6KgQJPreVisqLVn224pXG98Avf4=; b=EjSTJMBHZt3hW5GiTQLc0yoaL4bRSSv++j6JII24ZBJdLw9hG4wTVqwgE0fBBSGsAN Cet/6xjibNoLJw4SVbv2yC5id3HzUSt6lQea04aQv+ycEyN+4aQHv5CrQcWChKuh0HRG zgR8JX3LUIEMVfctCohEelCIdGGLpD0L+uxW0=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:cc:in-reply-to:mime-version :content-type:content-transfer-encoding:content-disposition :references; b=PW6vRYBaXUmZZlYxobcboQAERNVV9U3AvtAOjOFrs6TFnmxobgBqy1pTTpHh8JnMHE bZS7A9Ud5lJBHAtuzB3KRrgPZ6vqF3jbSzJ9qBOjF7JJa7tH2K3X13bN3pk3VDjhjU/j CQvYWThodt9/4JTQmsNS7lUNGFL27pufL8PQQ=
- References: <200806301747.07820.franke.daniel@gmail.com>
On Mon, Jun 30, 2008 at 5:47 PM, Daniel Franke <franke.daniel@gmail.com> wrote:
>
> Hi all.
>
> Is anybody aware of regressions in the optimizer on x86_64?
>
> With latest SVN, my code runs nicely up to -O1, but goes haywire on -O2. This
> is not the case for 4.3.2 (both checkouts from 20080629).
>
> Any pointers to PRs or how to (easily) track down the offending option and
> code are be highly welcome. I think I did not go for optimizer problems
> before ...
Instead of searching for the optimization flag causing this I would suggest
to run the code through a profiler to spot the point in the source
that regresses
in performance. It should then be not too difficult to compare the
optimized tree-dumps or the assembly of the regressed part - which is
also way more useful than a bugreport stating that option -ffoo makes things
worse.
Thanks,
Richard.