This is the mail archive of the gcc@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: Acovea 4, with Pentium 4 and Opteron results


Scott Robert Ladd <coyote@coyotegulch.com> writes:
>
> For the Opteron, Acovea produced clear results, improving the overall
> execution time of the benchmark suite by around 10% over the same code
> compiled with -O3 -ffast-math.

for amd64: 

alma
Strong optimisms: -fschedule-insns -mieee-fp -funsafe-math-optimizations
Strong pessimisms: -ffloat-store -mfpmath=sse

this sounds broken: -mfpmath=sse is the default, so why would the 
flag make it worse?

-Andi


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