1. 25 Mar, 2016 1 commit
    • Simon Peyton Jones's avatar
      Do not eta-reduce across Ticks in CorePrep · 356e5e03
      Simon Peyton Jones authored
      The function tryEtaReducePrep was being over-ambitious.
      When Breakpoint ticks were involved (i.e. in GHCi), eta
      reduction left an out-of-scope variable in the Tick.
      
      Easily fixed. Fixes the original report in Trac #111728.
      356e5e03
  2. 24 Mar, 2016 1 commit
    • Ben Gamari's avatar
      ErrUtils: Add timings to compiler phases · 8048d51b
      Ben Gamari authored
      This adds timings and allocation figures to the compiler's output when
      run with `-v2` in an effort to ease performance analysis.
      
      Todo:
        * Documentation
        * Where else should we add these?
        * Perhaps we should remove some of the now-arguably-redundant
          `showPass` occurrences where they are
        * Must we force more?
        * Perhaps we should place this behind a `-ftimings` instead of `-v2`
      
      Test Plan: `ghc -v2 Test.hs`, look at the output
      
      Reviewers: hvr, goldfire, simonmar, austin
      
      Reviewed By: simonmar
      
      Subscribers: angerman, michalt, niteria, ezyang, thomie
      
      Differential Revision: https://phabricator.haskell.org/D1959
      8048d51b
  3. 09 Mar, 2016 1 commit
    • Simon Peyton Jones's avatar
      Fix the implementation of lazyId · 4c3a0a4a
      Simon Peyton Jones authored
      'lazy' was doing part of its job, but not all!  In particular,
      an application
        f (lazy e)
      where f is strict, was still being compiled using call-by-value in
      CorePrep.  This defeated the purpose of defining catch as
         catch a b = catch# (lazy a) b
      See Trac #11555, and Neil Mitchell's test case in comment:14
      
      This patch makes 'lazy' behave properly. I updated Note [lazyId magic]
      in MkId, but all the action is in CorePrep.
      
      I can't say I really like this, but it does the job.
      4c3a0a4a
  4. 24 Feb, 2016 1 commit
    • eir@cis.upenn.edu's avatar
      Address #11471 by putting RuntimeRep in kinds. · d8c64e86
      eir@cis.upenn.edu authored
      See Note [TYPE] in TysPrim. There are still some outstanding
      pieces in #11471 though, so this doesn't actually nail the bug.
      
      This commit also contains a few performance improvements:
      
      * Short-cut equality checking of nullary type syns
      
      * Compare types before kinds in eqType
      
      * INLINE coreViewOneStarKind
      
      * Store tycon binders separately from kinds.
      
      This resulted in a ~10% performance improvement in compiling
      the Cabal package. No change in functionality other than
      performance. (This affects the interface file format, though.)
      
      This commit updates the haddock submodule.
      d8c64e86
  5. 27 Jan, 2016 1 commit
  6. 18 Jan, 2016 1 commit
    • Jan Stolarek's avatar
      Replace calls to `ptext . sLit` with `text` · b8abd852
      Jan Stolarek authored
      Summary:
      In the past the canonical way for constructing an SDoc string literal was the
      composition `ptext . sLit`.  But for some time now we have function `text` that
      does the same.  Plus it has some rules that optimize its runtime behaviour.
      This patch takes all uses of `ptext . sLit` in the compiler and replaces them
      with calls to `text`.  The main benefits of this patch are clener (shorter) code
      and less dependencies between module, because many modules now do not need to
      import `FastString`.  I don't expect any performance benefits - we mostly use
      SDocs to report errors and it seems there is little to be gained here.
      
      Test Plan: ./validate
      
      Reviewers: bgamari, austin, goldfire, hvr, alanz
      
      Subscribers: goldfire, thomie, mpickering
      
      Differential Revision: https://phabricator.haskell.org/D1784
      b8abd852
  7. 01 Jan, 2016 1 commit
    • Simon Peyton Jones's avatar
      Add strictness for runRW# · f3cc3456
      Simon Peyton Jones authored
      runRW# isn't inlined until CorePrep, so it's good to expose its
      strictness.  Moreover, if we don't we can get obscure failures
      in coreToStg; see Note [runRW arg] in CorePrep.
      
      This fixes Trac #11291, and makes DfltProb1 compile with -O
      always in order to expose it more vigorously
      f3cc3456
  8. 31 Dec, 2015 1 commit
  9. 24 Dec, 2015 1 commit
  10. 11 Dec, 2015 1 commit
    • eir@cis.upenn.edu's avatar
      Add kind equalities to GHC. · 67465497
      eir@cis.upenn.edu authored
      This implements the ideas originally put forward in
      "System FC with Explicit Kind Equality" (ICFP'13).
      
      There are several noteworthy changes with this patch:
       * We now have casts in types. These change the kind
         of a type. See new constructor `CastTy`.
      
       * All types and all constructors can be promoted.
         This includes GADT constructors. GADT pattern matches
         take place in type family equations. In Core,
         types can now be applied to coercions via the
         `CoercionTy` constructor.
      
       * Coercions can now be heterogeneous, relating types
         of different kinds. A coercion proving `t1 :: k1 ~ t2 :: k2`
         proves both that `t1` and `t2` are the same and also that
         `k1` and `k2` are the same.
      
       * The `Coercion` type has been significantly enhanced.
         The documentation in `docs/core-spec/core-spec.pdf` reflects
         the new reality.
      
       * The type of `*` is now `*`. No more `BOX`.
      
       * Users can write explicit kind variables in their code,
         anywhere they can write type variables. For backward compatibility,
         automatic inference of kind-variable binding is still permitted.
      
       * The new extension `TypeInType` turns on the new user-facing
         features.
      
       * Type families and synonyms are now promoted to kinds. This causes
         trouble with parsing `*`, leading to the somewhat awkward new
         `HsAppsTy` constructor for `HsType`. This is dispatched with in
         the renamer, where the kind `*` can be told apart from a
         type-level multiplication operator. Without `-XTypeInType` the
         old behavior persists. With `-XTypeInType`, you need to import
         `Data.Kind` to get `*`, also known as `Type`.
      
       * The kind-checking algorithms in TcHsType have been significantly
         rewritten to allow for enhanced kinds.
      
       * The new features are still quite experimental and may be in flux.
      
       * TODO: Several open tickets: #11195, #11196, #11197, #11198, #11203.
      
       * TODO: Update user manual.
      
      Tickets addressed: #9017, #9173, #7961, #10524, #8566, #11142.
      Updates Haddock submodule.
      67465497
  11. 04 Dec, 2015 1 commit
    • Simon Peyton Jones's avatar
      Case-of-empty-alts is trivial (Trac #11155) · 1c9fd3f1
      Simon Peyton Jones authored
      As you'll see from Trac #11155, the code generator was confused
      by a binding let x = y in ....   Why did that happen? Because of
      a (case y of {}) expression on the RHS.
      
      The right thing is just to expand what a "trivial" expression is.
      
      See Note [Empty case is trivial] in CoreUtils.
      1c9fd3f1
  12. 23 Nov, 2015 1 commit
  13. 12 Nov, 2015 1 commit
  14. 15 Oct, 2015 1 commit
  15. 10 Oct, 2015 1 commit
  16. 30 Aug, 2015 1 commit
  17. 26 Aug, 2015 1 commit
  18. 31 Mar, 2015 1 commit
    • Herbert Valerio Riedel's avatar
      Drop old integer-gmp-0.5 from GHC source tree · 995e8c1c
      Herbert Valerio Riedel authored
      This completes what c774b28f (#9281)
      started.  `integer-gmp-1.0` was added as an additional
      `libraries/integer-gmp2` folder while retaining the ability to configure
      GHC w/ the old `integer-gmp-0.5` to have a way back, and or the ability
      to easily switch between old/new `integer-gmp` for benchmark/debugging
      purposes.
      
      This commit removes the old `libraries/integer-gmp` folder and moves
      `libraries/integer-gmp2` into its place, while removing any mentions of
      "gmp2" as well as the to support two different `integer-gmp` packages in
      GHC's source-tree.
      
      Reviewed By: austin
      
      Differential Revision: https://phabricator.haskell.org/D769
      995e8c1c
  19. 10 Feb, 2015 1 commit
  20. 17 Dec, 2014 1 commit
    • Simon Peyton Jones's avatar
      Improve TidyPgm.hasCafRefs to account for Integer literals (Trac #8525) · 6b11bab6
      Simon Peyton Jones authored
      See Note [Disgusting computation of CafRefs] in TidyPgm.
      
      Also affects CoreUtils.rhsIsStatic.
      
      The real solution here is to compute CAF and arity information
      from the STG-program, and feed it back to tidied program for
      the interface file and later GHCi clients.  A battle for another
      day.
      
      But at least this commit reduces the number of gratuitous CAFs, and
      hence SRT entries.  And kills off a batch of ASSERT failures.
      6b11bab6
  21. 16 Dec, 2014 2 commits
    • Peter Wortmann's avatar
      Source notes (CorePrep and Stg support) · 4cdbf802
      Peter Wortmann authored
      This is basically just about continuing maintaining source notes after
      the Core stage. Unfortunately, this is more involved as it might seem,
      as there are more restrictions on where ticks are allowed to show up.
      
      Notes:
      
      * We replace the StgTick / StgSCC constructors with a unified StgTick
        that can carry any tickish.
      
      * For handling constructor or lambda applications, we generally float
        ticks out.
      
      * Note that thanks to the NonLam placement, we know that source notes
        can never appear on lambdas. This means that as long as we are
        careful to always use mkTick, we will never violate CorePrep
        invariants.
      
      * This is however not automatically true for eta expansion, which
        needs to somewhat awkwardly strip, then re-tick the expression in
        question.
      
      * Where CorePrep floats out lets, we make sure to wrap them in the
        same spirit as FloatOut.
      
      * Detecting selector thunks becomes a bit more involved, as we can run
        into ticks at multiple points.
      
      (From Phabricator D169)
      4cdbf802
    • Peter Wortmann's avatar
      Source notes (Core support) · 993975d3
      Peter Wortmann authored
      This patch introduces "SourceNote" tickishs that link Core to the
      source code that generated it. The idea is to retain these source code
      links throughout code transformations so we can eventually relate
      object code all the way back to the original source (which we can,
      say, encode as DWARF information to allow debugging).  We generate
      these SourceNotes like other tickshs in the desugaring phase. The
      activating command line flag is "-g", consistent with the flag other
      compilers use to decide DWARF generation.
      
      Keeping ticks from getting into the way of Core transformations is
      tricky, but doable. The changes in this patch produce identical Core
      in all cases I tested -- which at this point is GHC, all libraries and
      nofib. Also note that this pass creates *lots* of tick nodes, which we
      reduce somewhat by removing duplicated and overlapping source
      ticks. This will still cause significant Tick "clumps" - a possible
      future optimization could be to make Tick carry a list of Tickishs
      instead of one at a time.
      
      (From Phabricator D169)
      993975d3
  22. 15 Dec, 2014 1 commit
    • Simon Peyton Jones's avatar
      Make Core Lint check for locally-bound GlobalIds · d59c59f4
      Simon Peyton Jones authored
      There should be no bindings in this module for a GlobalId;
      except after CoreTidy, when top-level bindings are globalised.
      
      To check for this, I had to make the CoreToDo pass part of the
      environment that Core Lint caries.  But CoreToDo is defined in
      CoreMonad, which (before this patch) called CoreLint.
      
      So I had to do quite a bit of refactoring, moving some
      lint-invoking code into CoreLint itself.  Crucially, I also
      more tcLookupImported_maybe, importDecl, and checkwiredInTyCon
      from TcIface (which use CoreLint) to LoadIface (which doesn't).
      This is probably better structure anyway.
      
      So most of this patch is refactoring. The actual check for
      GlobalIds is in CoreLint.lintAndScopeId
      d59c59f4
  23. 03 Dec, 2014 1 commit
  24. 12 Nov, 2014 1 commit
    • Herbert Valerio Riedel's avatar
      Implement new integer-gmp2 from scratch (re #9281) · c774b28f
      Herbert Valerio Riedel authored
      This is done as a separate `integer-gmp2` backend library because it
      turned out to become a complete rewrite from scratch.
      
      Due to the different (over)allocation scheme and potentially different
      accounting (via the new `{shrink,resize}MutableByteArray#` primitives),
      some of the nofib benchmarks actually results in increased allocation
      numbers (but not necessarily an increase in runtime!).  I believe the
      allocation numbers could improve if `{resize,shrink}MutableByteArray#`
      could be optimised to reallocate in-place more efficiently.
      
      Here are the more apparent changes in the latest nofib comparision
      between `integer-gmp` and `integer-gmp2`:
      
        ------------------------------------------------------------------
                Program     Size    Allocs   Runtime   Elapsed  TotalMem
        ------------------------------------------------------------------
                    ...
             bernouilli    +1.6%    +15.3%     0.132     0.132      0.0%
                    ...
           cryptarithm1    -2.2%      0.0%     -9.7%     -9.7%      0.0%
                    ...
                  fasta    -0.7%     -0.0%    +10.9%    +10.9%      0.0%
                    ...
                  kahan    +0.6%    +38.9%     0.169     0.169      0.0%
                    ...
                   lcss    -0.7%     -0.0%     -6.4%     -6.4%      0.0%
                    ...
                 mandel    +1.6%    +33.6%     0.049     0.049      0.0%
                    ...
               pidigits    +0.8%     +8.5%     +3.9%     +3.9%      0.0%
                  power    +1.4%    -23.8%    -18.6%    -18.6%    -16.7%
                    ...
              primetest    +1.3%    +50.1%     0.085     0.085      0.0%
                    ...
                    rsa    +1.6%    +53.4%     0.026     0.026      0.0%
                    ...
                    scs    +1.2%     +6.6%     +6.5%     +6.6%    +14.3%
                    ...
                 symalg    +1.0%     +9.5%     0.010     0.010      0.0%
                    ...
              transform    -0.6%     -0.0%     -5.9%     -5.9%      0.0%
                    ...
        ------------------------------------------------------------------
                    Min    -2.3%    -23.8%    -18.6%    -18.6%    -16.7%
                    Max    +1.6%    +53.4%    +10.9%    +10.9%    +14.3%
         Geometric Mean    -0.3%     +1.9%     -0.8%     -0.8%     +0.0%
      
      (see P35 / https://phabricator.haskell.org/P35 for full report)
      
      By default, `INTEGER_LIBRARY=integer-gmp2` is active now, which results
      in the package `integer-gmp-1.0.0.0` being registered in the package db.
      The previous `integer-gmp-0.5.1.0` can be restored by setting
      `INTEGER_LIBRARY=integer-gmp` (but will probably be removed altogether
      for GHC 7.12). In-tree GMP support has been stolen from the old
      `integer-gmp` (while unpatching the custom memory-allocators, as well as
      forcing `-fPIC`)
      
      A minor hack to `ghc-cabal` was necessary in order to support two different
      `integer-gmp` packages (in different folders) with the same package key.
      
      There will be a couple of follow-up commits re-implementing some features
      that were dropped to keep D82 minimal, as well as further
      clean-ups/improvements.
      
      More information can be found via #9281 and
      https://ghc.haskell.org/trac/ghc/wiki/Design/IntegerGmp2
      
      Reviewed By: austin, rwbarton, simonmar
      
      Differential Revision: https://phabricator.haskell.org/D82
      c774b28f
  25. 04 Nov, 2014 1 commit
  26. 27 Oct, 2014 1 commit
    • Herbert Valerio Riedel's avatar
      Un-wire `Integer` type (re #9714) · 0e1f0f7d
      Herbert Valerio Riedel authored
      Integer is currently a wired-in type for integer-gmp. This requires
      replicating its inner structure in `TysWiredIn`, which makes it much
      harder to change Integer to a more complex representation (as
      e.g. needed for implementing #9281)
      
      This commit stops `Integer` being a wired-in type, and makes it
      known-key type instead, thereby simplifying code notably.
      
      Reviewed By: austin
      
      Differential Revision: https://phabricator.haskell.org/D351
      0e1f0f7d
  27. 21 Jul, 2014 1 commit
    • Edward Z. Yang's avatar
      Rename PackageId to PackageKey, distinguishing it from Cabal's PackageId. · 4bebab25
      Edward Z. Yang authored
      Summary:
      Previously, both Cabal and GHC defined the type PackageId, and we expected
      them to be roughly equivalent (but represented differently).  This refactoring
      separates these two notions.
      
      A package ID is a user-visible identifier; it's the thing you write in a
      Cabal file, e.g. containers-0.9.  The components of this ID are semantically
      meaningful, and decompose into a package name and a package vrsion.
      
      A package key is an opaque identifier used by GHC to generate linking symbols.
      Presently, it just consists of a package name and a package version, but
      pursuant to #9265 we are planning to extend it to record other information.
      Within a single executable, it uniquely identifies a package.  It is *not* an
      InstalledPackageId, as the choice of a package key affects the ABI of a package
      (whereas an InstalledPackageId is computed after compilation.)  Cabal computes
      a package key for the package and passes it to GHC using -package-name (now
      *extremely* misnamed).
      
      As an added bonus, we don't have to worry about shadowing anymore.
      
      As a follow on, we should introduce -current-package-key having the same role as
      -package-name, and deprecate the old flag.  This commit is just renaming.
      
      The haddock submodule needed to be updated.
      Signed-off-by: default avatarEdward Z. Yang <ezyang@cs.stanford.edu>
      
      Test Plan: validate
      
      Reviewers: simonpj, simonmar, hvr, austin
      
      Subscribers: simonmar, relrod, carter
      
      Differential Revision: https://phabricator.haskell.org/D79
      
      Conflicts:
      	compiler/main/HscTypes.lhs
      	compiler/main/Packages.lhs
      	utils/haddock
      4bebab25
  28. 27 May, 2014 1 commit
  29. 15 May, 2014 1 commit
    • Herbert Valerio Riedel's avatar
      Add LANGUAGE pragmas to compiler/ source files · 23892440
      Herbert Valerio Riedel authored
      In some cases, the layout of the LANGUAGE/OPTIONS_GHC lines has been
      reorganized, while following the convention, to
      
      - place `{-# LANGUAGE #-}` pragmas at the top of the source file, before
        any `{-# OPTIONS_GHC #-}`-lines.
      
      - Moreover, if the list of language extensions fit into a single
        `{-# LANGUAGE ... -#}`-line (shorter than 80 characters), keep it on one
        line. Otherwise split into `{-# LANGUAGE ... -#}`-lines for each
        individual language extension. In both cases, try to keep the
        enumeration alphabetically ordered.
        (The latter layout is preferable as it's more diff-friendly)
      
      While at it, this also replaces obsolete `{-# OPTIONS ... #-}` pragma
      occurences by `{-# OPTIONS_GHC ... #-}` pragmas.
      23892440
  30. 21 Nov, 2013 1 commit
  31. 03 Sep, 2013 1 commit
    • Simon Peyton Jones's avatar
      Improve Linting in GHCi (fixes Trac #8215) · dfa8ef03
      Simon Peyton Jones authored
      The original problem was that we weren't bringing varaibles bound in the
      interactive context into scope before Linting the result of a top-level
      declaration in GHCi.  (We were doing this for expressions.)
      
      Moreover I found that we weren't Linting the result of desugaring
      a GHCi expression, which we really should be doing.
      
      It took me a bit of time to unravel all this, and I did some refactoring
      to make it easier next time.
      
        * CoreMonad contains the Lint wrappers that get the right
          environments into place.  It always had endPass and lintPassResult
          (which Lints bindings), but now it has lintInteractiveExpr.
      
        * Both use a common function CoreMonad.interactiveInScope to find
          those in-scope variables.
      
      Quite a bit of knock-on effects from this, but nothing exciting.
      dfa8ef03
  32. 06 Jun, 2013 1 commit
    • Simon Peyton Jones's avatar
      Implement cardinality analysis · 99d4e5b4
      Simon Peyton Jones authored
      This major patch implements the cardinality analysis described
      in our paper "Higher order cardinality analysis". It is joint
      work with Ilya Sergey and Dimitrios Vytiniotis.
      
      The basic is augment the absence-analysis part of the demand
      analyser so that it can tell when something is used
      	 never
      	 at most once
       	 some other way
      
      The "at most once" information is used
          a) to enable transformations, and
             in particular to identify one-shot lambdas
          b) to allow updates on thunks to be omitted.
      
      There are two new flags, mainly there so you can do performance
      comparisons:
          -fkill-absence   stops GHC doing absence analysis at all
          -fkill-one-shot  stops GHC spotting one-shot lambdas
                           and single-entry thunks
      
      The big changes are:
      
      * The Demand type is substantially refactored.  In particular
        the UseDmd is factored as follows
            data UseDmd
              = UCall Count UseDmd
              | UProd [MaybeUsed]
              | UHead
              | Used
      
            data MaybeUsed = Abs | Use Count UseDmd
      
            data Count = One | Many
      
        Notice that UCall recurses straight to UseDmd, whereas
        UProd goes via MaybeUsed.
      
        The "Count" embodies the "at most once" or "many" idea.
      
      * The demand analyser itself was refactored a lot
      
      * The previously ad-hoc stuff in the occurrence analyser for foldr and
        build goes away entirely.  Before if we had build (\cn -> ...x... )
        then the "\cn" was hackily made one-shot (by spotting 'build' as
        special.  That's essential to allow x to be inlined.  Now the
        occurrence analyser propagates info gotten from 'build's stricness
        signature (so build isn't special); and that strictness sig is
        in turn derived entirely automatically.  Much nicer!
      
      * The ticky stuff is improved to count single-entry thunks separately.
      
      One shortcoming is that there is no DEBUG way to spot if an
      allegedly-single-entry thunk is acually entered more than once.  It
      would not be hard to generate a bit of code to check for this, and it
      would be reassuring.  But it's fiddly and I have not done it.
      
      Despite all this fuss, the performance numbers are rather under-whelming.
      See the paper for more discussion.
      
             nucleic2          -0.8%    -10.9%      0.10      0.10     +0.0%
               sphere          -0.7%     -1.5%      0.08      0.08     +0.0%
      --------------------------------------------------------------------------------
                  Min          -4.7%    -10.9%     -9.3%     -9.3%    -50.0%
                  Max          -0.4%     +0.5%     +2.2%     +2.3%     +7.4%
       Geometric Mean          -0.8%     -0.2%     -1.3%     -1.3%     -1.8%
      
      I don't quite know how much credence to place in the runtime changes,
      but movement seems generally in the right direction.
      99d4e5b4
  33. 19 May, 2013 2 commits
  34. 28 Mar, 2013 1 commit
  35. 27 Mar, 2013 1 commit
  36. 17 Jan, 2013 1 commit
    • Simon Peyton Jones's avatar
      Major patch to implement the new Demand Analyser · 0831a12e
      Simon Peyton Jones authored
      This patch is the result of Ilya Sergey's internship at MSR.  It
      constitutes a thorough overhaul and simplification of the demand
      analyser.  It makes a solid foundation on which we can now build.
      Main changes are
      
      * Instead of having one combined type for Demand, a Demand is
         now a pair (JointDmd) of
            - a StrDmd and
            - an AbsDmd.
         This allows strictness and absence to be though about quite
         orthogonally, and greatly reduces brain melt-down.
      
      * Similarly in the DmdResult type, it's a pair of
           - a PureResult (indicating only divergence/non-divergence)
           - a CPRResult (which deals only with the CPR property
      
      * In IdInfo, the
          strictnessInfo field contains a StrictSig, not a Maybe StrictSig
          demandInfo     field contains a Demand, not a Maybe Demand
        We don't need Nothing (to indicate no strictness/demand info)
        any more; topSig/topDmd will do.
      
      * Remove "boxity" analysis entirely.  This was an attempt to
        avoid "reboxing", but it added complexity, is extremely
        ad-hoc, and makes very little difference in practice.
      
      * Remove the "unboxing strategy" computation. This was an an
        attempt to ensure that a worker didn't get zillions of
        arguments by unboxing big tuples.  But in fact removing it
        DRAMATICALLY reduces allocation in an inner loop of the
        I/O library (where the threshold argument-count had been
        set just too low).  It's exceptional to have a zillion arguments
        and I don't think it's worth the complexity, especially since
        it turned out to have a serious performance hit.
      
      * Remove quite a bit of ad-hoc cruft
      
      * Move worthSplittingFun, worthSplittingThunk from WorkWrap to
        Demand. This allows JointDmd to be fully abstract, examined
        only inside Demand.
      
      Everything else really follows from these changes.
      
      All of this is really just refactoring, so we don't expect
      big performance changes, but acutally the numbers look quite
      good.  Here is a full nofib run with some highlights identified:
      
              Program           Size    Allocs   Runtime   Elapsed  TotalMem
      --------------------------------------------------------------------------------
               expert          -2.6%    -15.5%      0.00      0.00     +0.0%
                fluid          -2.4%     -7.1%      0.01      0.01     +0.0%
                   gg          -2.5%    -28.9%      0.02      0.02    -33.3%
            integrate          -2.6%     +3.2%     +2.6%     +2.6%     +0.0%
              mandel2          -2.6%     +4.2%      0.01      0.01     +0.0%
             nucleic2          -2.0%    -16.3%      0.11      0.11     +0.0%
                 para          -2.6%    -20.0%    -11.8%    -11.7%     +0.0%
               parser          -2.5%    -17.9%      0.05      0.05     +0.0%
               prolog          -2.6%    -13.0%      0.00      0.00     +0.0%
               puzzle          -2.6%     +2.2%     +0.8%     +0.8%     +0.0%
              sorting          -2.6%    -35.9%      0.00      0.00     +0.0%
             treejoin          -2.6%    -52.2%     -9.8%     -9.9%     +0.0%
      --------------------------------------------------------------------------------
                  Min          -2.7%    -52.2%    -11.8%    -11.7%    -33.3%
                  Max          -1.8%     +4.2%    +10.5%    +10.5%     +7.7%
       Geometric Mean          -2.5%     -2.8%     -0.4%     -0.5%     -0.4%
      
      Things to note
      
      * Binary sizes are smaller. I don't know why, but it's good.
      
      * Allocation is sometiemes a *lot* smaller. I believe that all the big numbers
        (I checked treejoin, gg, sorting) arise from one place, namely a function
        GHC.IO.Encoding.UTF8.utf8_decode, which is strict in two Buffers both of
        which have several arugments.  Not w/w'ing both arguments (which is what
        we did before) has a big effect.  So the big win in actually somewhat
        accidental, gained by removing the "unboxing strategy" code.
      
      * A couple of benchmarks allocate slightly more.  This turns out
        to be due to reboxing (integrate).  But the biggest increase is
        mandel2, and *that* turned out also to be a somewhat accidental
        loss of CSE, and pointed the way to doing better CSE: see Trac
        #7596.
      
      * Runtimes are never very reliable, but seem to improve very slightly.
      
      All in all, a good piece of work.  Thank you Ilya!
      0831a12e
  37. 17 Sep, 2012 1 commit
  38. 28 Aug, 2012 1 commit