This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/42216] [4.5 Regression] rev 15458[78] regress 464.h264ref peak 20%
- From: "rguenther at suse dot de" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 1 Dec 2009 15:35:49 -0000
- Subject: [Bug tree-optimization/42216] [4.5 Regression] rev 15458[78] regress 464.h264ref peak 20%
- References: <bug-42216-10053@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #6 from rguenther at suse dot de 2009-12-01 15:35 -------
Subject: Re: [4.5 Regression] rev 15458[78]
regress 464.h264ref peak 20%
On Tue, 1 Dec 2009, bernds_cb1 at t-online dot de wrote:
> ------- Comment #5 from bernds_cb1 at t-online dot de 2009-12-01 15:26 -------
> Code generation changes are expected for two reasons - the code became less
> conservative when determining conflicts with other registers, so we can usually
> rename in more cases. There are also a few cases where we have to give up on
> renaming due to multiword hard register overlaps we can't track.
>
> Unfortunately I don't have access to spec; I'll have to think about how to
> track this down.
I will try to come up with a testcase for you. I suppose it might
be a backend issue as regrename should mainly effect the 2nd
scheduling pass, right?
Richard.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42216