nofib regressions in HEAD since 7.6.2 release

Johan Tibell johan.tibell at gmail.com
Thu Feb 7 23:04:04 CET 2013


And here are the shootout benchmark results:

--------------------------------------------------------------------------------
        Program           Size    Allocs   Runtime   Elapsed  TotalMem
--------------------------------------------------------------------------------
   binary-trees          -6.4%     +0.0%     -2.4%     +0.0%     -7.4%
 fannkuch-redux          -0.8%     +0.0%     -3.6%     -3.6%     +0.0%
          fasta          -0.8%     +0.0%     -4.8%     -0.6%     +0.0%
         n-body          -6.2%     -0.0%    +28.5%    +29.4%     +0.0%
       pidigits          -0.8%     +0.0%     -2.9%     +2.9%     +0.0%
  spectral-norm          -6.3%    -13.9%     -0.0%     -0.0%     +0.0%
--------------------------------------------------------------------------------
            Min          -6.4%    -13.9%     -4.8%     -3.6%     -7.4%
            Max          -0.8%     +0.0%    +28.5%    +29.4%     +0.0%
 Geometric Mean          -3.6%     -2.5%     +1.9%     +4.2%     -1.3%

It might be worth looking into why n-body is 28% slower.

P.S. I've now enabled the shootout benchmarks in nofib by default, as
they're pretty quick to run and run on any machine.

-- Johan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.haskell.org/pipermail/ghc-devs/attachments/20130207/8bd94a2e/attachment.htm>


More information about the ghc-devs mailing list