This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: x86_64 varargs setup jump table

On 07/22/2010 11:02 AM, Sebastian Pop wrote:
> Here are the results on AMD Phenom(tm) 9950 Quad-Core.
> 
> Old: Gcc 4.6.0 revision 162355
> New: Gcc 4.6.0 revision 162355 + this patch.
> Flags: -O3 -funroll-loops -fpeel-loops -ffast-math -march=native
> 
> The number is the run time percentage: (old - new) / old * 100
> (positive is better)
> 
> [ no positive results ]

Hmm.  At least HJ had some positive results.  I'm surprised
that there are none on the AMD box.

Does movaps have reformatting stalls that perhaps movdqa does
with that particular micro-architecture?  Or are stores exempt
from reformatting stalls now?

Otherwise the only thing I can think is that the computed jump
was in practice very predictable (i.e. lots of calls containing
the same sequence of types), and that performing a few less
stores makes that difference.


r~
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]