This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Improve heuristic for global-vs-local regs
- From: Adrian Straetling <straetling at de dot ibm dot com>
- To: Michael Matz <matz at suse dot de>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Thu, 1 Dec 2005 18:47:41 +0100
- Subject: Re: Improve heuristic for global-vs-local regs
- References: <Pine.LNX.4.58.0512010454170.7673@wotan.suse.de>
On Thu, Dec 01, 2005 at 05:11:10AM +0100, Michael Matz wrote:
> Hi,
>
> while investigating some performance problems in lucas on x86-64 I noticed
> that one hot loop was calculating only with memory operands, although
> enough regs should have been free. It's an explicitely unrolled
> 16-butterfly of an FFT, so it's first loading 15 values into 16 SSE
> registers, then _spills all of them to stack_, and henceforth happily does
> this:
...
Your patch also improves lucas on s390x by approximately 15%.
bye,
Adrian