Skip to content
Commits on Oct 8, 2010
Commits on Oct 7, 2010
  1. Don't normalize by default for fibon-analyse raw output

    committed
    The raw output is intended to be input into another tool, such as R,
    for further processing so don't normalize the data by default. Can
    still be overridedn by passing a normalize flag.
  2. Merge branch 'stats_parser'

    committed
Commits on Oct 5, 2010
  1. Refactor analysis routines into separate module

    committed
    The analysis routines for creating the fibon data and extraStats
    data have been moved to their own module. We should build up a set
    of analysis routines that use the resampling technique similar to
    Criterion.
Commits on Oct 4, 2010
Commits on Oct 3, 2010
  1. Fix bug in finding benchmarks on windows

    committed
    The benchmarks were getting added to the wrong group which then
    caused a compile error because the Instance.hs file could not be
    found. The fix is to sort the groups first, and use the sorted
    groups consistently throughtout the function.
Commits on Oct 2, 2010
Commits on Oct 1, 2010
  1. Updated README

    committed
  2. Prepare for cabal upload

    committed
Commits on Sep 30, 2010
Commits on Sep 29, 2010
  1. Convert table generation to use normalization

    committed
    Use the new analysis routines to normalize the results to a baseline.
Commits on Sep 28, 2010
  1. Refactored metrics for new table generation code

    committed
    Getting ready to switch the table generation over to using the new
    analysis routines for normalizing data.
Commits on Sep 25, 2010
  1. Add missing dependencies for fibon-analyse

    committed
    Needed the depedencies added for fibon-analyse that were previously
    coming from other executable sections.
  2. Add flag to conditionally build fibon-analyse

    committed
    To support this we bumped Cabal dependency to 1.8. Versions before
    1.8 did not properly seperate out the dependencies in the executable
    sections. This means that all the build-depends were shared among
    all the executables. See http://hackage.haskell.org/trac/hackage/ticket/742
    
    With the later version of cabal we can conditionally build the
    analyse program by wrapping its build-depends in an if condition.
Commits on Sep 24, 2010
  1. Add more type info to PerfData

    committed
    We can be more explicit about what kind of performance data we are
    carrying around. Hopefully this makes it easier to write correct
    summary and normalization code.
  2. Track changes in config submodule

    committed
    * config 3dff706...5424222 (1):
      > Small changes for running with GHC 7
  3. Track changes in benchmarks submodule

    committed
    * benchmarks/Fibon/Benchmarks f10639a...26502d2 (2):
      > Add local type signatures for GHC 7
      > Fixed sloc to also work on nofib
Commits on Sep 23, 2010
  1. Rename MetricFormat to PerfData

    committed
  2. Allow fibon to be built without any benchmarks

    committed
    We can still build the fibon toolset without any benchmarks. We just
    emit a warning and generate some placeholder code in Benchmarks.hs
Commits on Sep 15, 2010
  1. Track changes in submodules

    committed
    Updated config and benchmarks for GHC Head
Commits on Sep 14, 2010
  1. Get the Shootout benchmarks to work with GHC 6.10

    committed
    Updated the config file and added some README notes for those
    benchmarks that don't work.
  2. Track submodules changes for new Shooutout benchmarks

    committed
    These changes add the shooutout benchmarks to the mix.
  3. Track changes in benchmark submodule

    committed
    The benchmarks were reorganized to all be in a Hackage group
  4. Track changes in config submodule

    committed
    Added new config files for HEAD and HEAD+LLVM.
Commits on Sep 13, 2010
  1. Added basic table generation for results.

    committed
    Currently only prints out raw results with no normilaztion or
    averaging.
Commits on Sep 11, 2010
Commits on Sep 10, 2010
Something went wrong with that request. Please try again.