Changes between Version 1 and Version 2 of Ticket #8457, comment 5


Ignore:
Timestamp:
Oct 20, 2013 8:35:34 PM (6 years ago)
Author:
errge
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #8457, comment 5

    v1 v2  
    55I followed http://ghc.haskell.org/trac/ghc/wiki/Building/RunningNoFib and the results are here: https://github.com/errge/notlazy/blob/master/nofib.txt
    66
    7 There are some outliers in both directions, but the overall picture seems to me as kind of "nothing changing".  Even more if you deduce the noise introduced by the outliers.
     7There are some outliers in both directions, but the overall picture seems to me as kind of "nothing changing".  Even more if you subtract the noise introduced by the outliers.
    88
    9 It may be meaningful to have a look on fulsom, hidden, parser and parstof; these are the examples where the optimization did very good and it would be interesting to see whether it's easy to reintroduce the necessary sharing by hand.  OTOH, I'm not really in the mood to debug thousand of lines of old school haskell written in '93 by academics. :-)  If we get really bored, maybe will take a look...
     9It may be worth to have a look on fulsom, hidden, parser and parstof; these are the examples where the optimization did very well and it would be interesting to see whether it's easy to reintroduce the necessary sharing by hand.  OTOH, these benchmarks are a bit old school, undocumented and some of them are autogenerated.
    1010
    1111And there are some outliers on the negative side: constraints, bspt and integer.  It'd be interesting to see here if the optimization introduced sharing or other performance bugs are easy to workaround or not in these cases.
    1212
    13 My opinion about turning this off didn't change, but I can always just turn off for myself, so feel free to either fix this or mark the bug as invalid.
     13And in general, we may have status quo bias here: when implementing and submitting these real world benchmarks people just used the GHC as it was at that point, optimizing for that.  If they were forced to do sharing then they would have.
    1414
    1515If there are more benchmarks to run, I'm happy to run them.