Permalink
Commits on Apr 20, 2018
  1. Use GHC 8.4.2 on Travis

    RyanGlScott committed Apr 20, 2018
Commits on Apr 7, 2018
Commits on Apr 5, 2018
  1. Add upper version bounds for base-compat

    RyanGlScott committed Apr 5, 2018
    Work around #192 until a new aeson release supporting
    base-compat-0.10 is out.
Commits on Mar 27, 2018
  1. Move measurement functionality to criterion-measurement (#191)

    ocramz authored and RyanGlScott committed Mar 27, 2018
    * FIXME module name overlap
    
    * fix imports
    
    * restore from deleting app/ by mistake
    
    * fix imports
    
    * fix Measurement interface
    
    * Update Travis to be criterion-measurement–aware
    
    * apply suggestions
    
    * apply suggestions
    
    * removed additional stack.yaml
    
    * oops
    
    * get rid of PackageImports
    
    * apply suggestions
    
    * apply suggestions
    
    * add README, apply suggestions
    
    * add criterion-measurement to examples/Overhead imports
    
    * update changelog.md
    
    * apply suggestions
    
    * apply suggestions
Commits on Mar 11, 2018
  1. Allow exceptions-0.10

    RyanGlScott committed Mar 11, 2018
Commits on Mar 9, 2018
Commits on Mar 4, 2018
  1. Version 1.4.0.0

    RyanGlScott committed Mar 4, 2018
Commits on Feb 27, 2018
  1. No need to hack around snoyberg/conduit#353 any longer

    RyanGlScott committed Feb 27, 2018
    Now that conduit-1.2.13.1 has been released
Commits on Feb 26, 2018
  1. Decrease impact of gc on benchmark performance (fixes #185) (#187)

    patrickdoc authored and RyanGlScott committed Feb 26, 2018
    * Remove extra performGC's, change the others to performMinorGC
    
    * Document performMinorGc in changelog
    
    * Add CPP guards to fix build on GHC 7.6 and 7.4
    
    * Fix formatting
    
    * Cleanup CPP and comment on performMinorGC
    
    * Fix performMinorGC foreign import
Commits on Feb 25, 2018
  1. Allow exceptions-0.9

    RyanGlScott committed Feb 25, 2018
Commits on Feb 15, 2018
  1. Bump the stack LTS version to 10.5 (#186)

    ckoparkar authored and RyanGlScott committed Feb 15, 2018
    It defaults to using optparse-applicative-0.14.0.0.
    So the `extra-dep` is unnecessary.
Commits on Feb 7, 2018
  1. Remove allocation from inner loops (fixes #183) (#184)

    patrickdoc authored and RyanGlScott committed Feb 7, 2018
    * Remove excess allocation
    
    * Refactor nf and whnf to reduce allocation
    
    * Refactor nfIO and whnfIO as well
    
    * Reorder exports
    
    * Adjust whnfIO' implementation to evaluate the result to WHNF, not the action
    
    * Adjust nfIO' and whnfIO' to be closer implementations
    
    * Add BangPatterns to Criterion.Types.Internal
    
    * Remove unused import: evaluate
    
    * Document changes
    
    * Document NOINLINE and fix typo
Commits on Feb 1, 2018
  1. Fix per-run combine time (fixes #181) (#182)

    patrickdoc authored and RyanGlScott committed Feb 1, 2018
    * Fix per-run combine time
    
    * Document change
Commits on Jan 18, 2018
  1. Addendum to #178

    RyanGlScott committed Jan 18, 2018
    [ci skip]
  2. Remove bad overhead code (#160, #161, #162) (#178)

    patrickdoc authored and RyanGlScott committed Jan 18, 2018
    * Remove bad overhead code, see #161
    
    * Update changelog
Commits on Jan 10, 2018
  1. performGC before getting RTS/GC stats (#177)

    avieth authored and RyanGlScott committed Jan 10, 2018
    * performGC before getting RTS/GC stats
    
    The deprecated getGCStats had some nice documentation: "If you would
    like your statistics as recent as possible, first run a performGC"
    
    This is sadly missing from getRTSStats but I believe it still holds.
    When regressing allocated over iters, I'd see a believable slope, but an
    unbelievable y intercept, something like -300000 when the slope is 150
    or so. That's because the first hundred or so measurements were 0 for
    bytes allocated, as the RTS didn't bother to run a GC before the RTS
    stats were sampled.
    
    Now we do 3 samples:
    
      1. performGC before the first one, to ensure it's up-to-date.
      2. Do the second one after the action, without a performGC, so we can
         get legit readings on the GC-related stats.
      3. performGC and then sample again, so we can get up-to-date readings
         on other metrics.
      4. Carefully choose whether to diff start stats against the end stats
         per- or post-GC.
    
    Also included is a fix to the ToJSON Measurements instance, which
    duplicated the mutator cpu seconds where GC cpu seconds should go.
    
    * fix shadowed binding warning
    
    * comment applyGCStatistics and simplify diff
    
    * update comment for getGCStatistics
Commits on Jan 9, 2018
  1. Version 1.3.0.0

    RyanGlScott committed Jan 9, 2018
  2. Mention #176 in changelog

    RyanGlScott committed Jan 9, 2018
  3. Fix GC/RTS stats for base >= 4.10.0 (#176)

    avieth authored and RyanGlScott committed Jan 9, 2018
    * Fix GC/RTS stats for base >= 4.10.0
    
    * correction: use gc_cpu_ns gc_elapsed_ns
Commits on Jan 7, 2018
  1. Old GHC fixes, pt. 3

    RyanGlScott committed Jan 7, 2018
  2. Depend on semigroups

    RyanGlScott committed Jan 7, 2018
  3. Old GHC fixes, pt. 2

    RyanGlScott committed Jan 7, 2018
  4. Old GHC fixes

    RyanGlScott committed Jan 7, 2018
  5. Tidy up the examples

    RyanGlScott committed Jan 7, 2018
  6. Fix some warnings

    RyanGlScott committed Jan 7, 2018
  7. More Travis tweaks

    RyanGlScott committed Jan 7, 2018