Polyhedron Fortran benchmark - compiler comparison: Good results for GCC 4.6

Steve Kargl sgk@troutmask.apl.washington.edu
Mon Sep 27 09:12:00 GMT 2010


On Sun, Sep 26, 2010 at 04:47:29PM +0200, Tobias Burnus wrote:
>  Hello all,
> 
> given the optimization work which has gone into 4.6, I thought I would 
> run the Polyhedron Fortran test suite.

I was interested in seeing how 4.6 stacks up against previous
versions of gfortran.  On x86_64-*-freebsd, using my preferred
set of options, I see

   Benchmark   gfc42    gfc43    gfc44    gfc45    gfc4x
        Name   (secs)   (secs)   (secs)   (secs)   (secs)
   ---------   -------  -------  -------  -------  -------
          ac    41.74    17.79    17.76    17.48    16.59
      aermod    44.66    45.70    47.21    46.23    41.41
         air    22.59    22.29    23.26    19.58    19.68
    capacita   116.02   117.91   112.57   116.89   116.43
     channel    20.52    20.56    20.33    20.83    20.83
       doduc    50.41    49.86    53.26    50.20    50.22 
     fatigue    21.86    16.66    14.69    12.48    13.18
     gas_dyn    20.12    16.16    16.06    15.37    14.42
      induct    65.57    62.25    49.10    32.59    32.82
       linpk    45.13    44.56    45.12    45.08    49.17
        mdbx    24.11    23.75    22.97    23.07    23.01
          nf    59.34    59.41    58.72    56.04    54.94
     protein    68.85    69.72    69.82    70.24    65.35
      rnflow    52.36    51.81    50.87    50.40    50.34
    test_fpu    32.66    33.66    33.89    33.69    33.92
        tfft    11.95    11.59    11.50    11.59    11.79

   Geo. Mean    37.16    34.04    33.33    31.57    31.28

options: -O2 -march=native -funroll-loops -ftree-vectorize

gfc42 --> gcc version 4.2.5 20090330
gfc43 --> gcc version 4.3.6 20100903
gfc44 --> gcc version 4.4.5 20100903
gfc45 --> gcc version 4.5.1 20100728
gfc4x --> gcc version 4.6.0 20100916

In general, I like the trend in the geometric mean.  The 
linpk tests looks suspicious but I've had no time to 
investiagte.

-- 
Steve



More information about the Gcc mailing list