[LLVMdev] x86_64-apple-darwin Polyhedron 2005 benchmarks
Chris Lattner
clattner at apple.com
Sun Aug 23 19:33:18 PDT 2009
On Aug 23, 2009, at 5:21 PM, Jack Howarth wrote:
> The current llvm/llvm-gcc-4.2 2.6 branch passes all of the
> Polyhedron 2005 benchmarks built with its gfortran. The
> results compare as follows...
Very interesting Jack! Do you have any idea why llvm-gcc is slower on
channel at -m64 than -m32? Can you attach some shark traces?
-Chris
>
> Compile Command : gfortran -ffast-math -funroll-loops -msse3 -O3
> %n.f90 -o %n
>
> benchmark gcc-4.2.4 llvm-gcc-svn llvm-gcc-2.6
> llvm-gcc-2.6
> at -m32 20081031 -m32 at -m32 at -
> m64
>
> ac 18.30 19.33 19.99 15.78
> aermod 30.90 22.02 23.52 20.84
> air 10.85 14.30 11.25 10.96
> capacita 63.58 63.65 62.15 54.38
> channel 6.29 4.68 4.00 4.29
> doduc 47.94 50.60 47.42 43.88
> fatigue 16.00 17.14 14.90 12.18
> gas_dyn 22.91 29.24 23.73 23.00
> induct 68.95 68.63 64.91 60.53
> linpk 31.06 30.77 26.67 26.60
> mdbx 14.94 18.30 17.62 15.44
> nf 33.98 35.01 35.41 38.72
> protein 51.03 55.97 57.65 48.73
> rnflow 43.20 47.68 40.09 39.75
> test_fpu 15.05 16.57 16.02 14.23
> tfft 2.94 2.87 2.86 2.87
>
> These benchmarks were all collected on a 2.33GHz MacBook Pro with
> 3Gb of ram.
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
More information about the llvm-dev
mailing list