This is the mail archive of the
mailing list for the GCC project.
Re: RFA: patch for PR37397
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: "Jeff Law" <law at redhat dot com>
- Cc: "Vladimir Makarov" <vmakarov at redhat dot com>, gcc-patches <gcc-patches at gcc dot gnu dot org>, "Kenneth Zadeck" <Kenneth dot Zadeck at naturalbridge dot com>
- Date: Mon, 17 Nov 2008 06:32:17 -0800
- Subject: Re: RFA: patch for PR37397
- References: <49187CDD.email@example.com> <491B0FB7.firstname.lastname@example.org>
On Wed, Nov 12, 2008 at 9:17 AM, Jeff Law <email@example.com> wrote:
> Vladimir Makarov wrote:
>> This patch improves SPEC benchmarks. I saw stable improvements on
>> x86/x86_64 and ppc. This patch implements a small trick mentioned in one
>> classical article by Chaitin etc (Register allocation and spilling via graph
>> coloring). There is no sense to spill pseudo in whose live range nothing is
>> dying because the spill will not make other allocnos colorable and
>> additional reloads for the corresponding pseudo will be generated in reload
>> pass for each insn it occurs.
>> Is it ok to commit?
> OK. Good find.
This caused 30% slowdown on 454.calculix in SPEC CPU 2006
with -O2 -ffast-math on Linux/Intel64.