[GHC] #16364: Derived Enum for small number of constructors seems suboptimal

26 Feb 2019

      #16364: Derived Enum for small number of constructors seems suboptimal
-------------------------------------+-------------------------------------
           Reporter:  Fuuzetsu       |             Owner:  (none)
               Type:  bug            |            Status:  new
           Priority:  normal         |         Milestone:
          Component:  Compiler       |           Version:  8.6.3
           Keywords:                 |  Operating System:  Unknown/Multiple
       Architecture:                 |   Type of failure:  None/Unknown
  Unknown/Multiple                   |
          Test Case:                 |        Blocked By:
           Blocking:                 |   Related Tickets:
Differential Rev(s):                 |         Wiki Page:
-------------------------------------+-------------------------------------
 {{{#!hs
 {-# LANGUAGE MagicHash #-}
 module En (toEnum', toEnum'', toEnum''', toEnum'''') where

 import GHC.Int
 import GHC.Prim
 import GHC.Word

 data Q' = Foo' | Bar'
     deriving Enum

 toEnum' :: Int -> Q'
 toEnum' 0 = Foo'
 toEnum' 1 = Bar'
 toEnum' x = error $ "out of range " <> show x

 toEnum'' :: Int -> Q'
 toEnum'' x@(I# n) | x >= 0 && x <= 1 = tagToEnum# n
 toEnum'' x = error $ "out of range " <> show x

 toEnum''' :: Int -> Q'
 toEnum''' x@(I# n) | x == 0 || x == 1 = tagToEnum# n
 toEnum''' x = error $ "out of range " <> show x

 toEnum'''' :: Int -> Q'
 toEnum'''' x@(I# n) = case int2Word# n `leWord#` 1## of
     0# -> error $ "out of range " <> show x
     _ -> tagToEnum# n
 }}}

 For the derived {{{toEnum}}}, we get something as
 {{{#!hs

 -- RHS size: {terms: 19, types: 4, coercions: 0, joins: 0/0}
 En.$w$ctoEnum [InlPrag=NOUSERINLINE[2]] :: Int# -> Q'
 [GblId,
  Arity=1,
  Str=,
  Unf=Unf{Src=<vanilla>, TopLvl=True, Value=True, ConLike=True,
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [0] 83 0}]
 En.$w$ctoEnum
   = \ (ww_s3Kl :: Int#) ->
       case >=# ww_s3Kl 0# of {
         __DEFAULT -> En.$wlvl ww_s3Kl;
         1# ->
           case <=# ww_s3Kl 1# of {
             __DEFAULT -> En.$wlvl ww_s3Kl;
             1# -> tagToEnum# @ Q' ww_s3Kl
           }
       }

 -- RHS size: {terms: 6, types: 3, coercions: 0, joins: 0/0}
 En.$fEnumQ'_$ctoEnum [InlPrag=NOUSERINLINE[2]] :: Int -> Q'
 [GblId,
  Arity=1,
  Str=,
  Unf=Unf{Src=InlineStable, TopLvl=True, Value=True, ConLike=True,
          WorkFree=True, Expandable=True,
          Guidance=ALWAYS_IF(arity=1,unsat_ok=True,boring_ok=False)
          Tmpl= \ (w_s3Ki [Occ=Once!] :: Int) ->
                  case w_s3Ki of { I# ww1_s3Kl [Occ=Once] ->
                  En.$w$ctoEnum ww1_s3Kl
                  }}]
 En.$fEnumQ'_$ctoEnum
   = \ (w_s3Ki :: Int) ->
       case w_s3Ki of { I# ww1_s3Kl -> En.$w$ctoEnum ww1_s3Kl }
 }}}

 Two comparisons to find out one thing! Contrast this with something like
 {{{toEnum'}}}:
 {{{#!hs
 toEnum'
   = \ (ds_d3dp :: Int) ->
       case ds_d3dp of { I# ds1_d3dr ->
       case ds1_d3dr of ds2_X3dR {
         __DEFAULT -> En.$wlvl1 ds2_X3dR;
         0# -> En.Foo';
         1# -> En.Bar'
       }
       }
 }}}

 Surely this seems better? But we don't even have to write out the
 constructors by hand in this case. {{{toEnum'''}}} actually produces the
 same code as {{{toEnum'}}}.

 I also wrote {{{toEnum''''}}} which I had some hopes for but actually runs
 the slowest. I'm unsure why. Seems simple enough:
 {{{#!hs
 toEnum''''
   = \ (x_a2Sd :: Int) ->
       case x_a2Sd of { I# n_a2Se ->
       case leWord# (int2Word# n_a2Se) 1## of {
         __DEFAULT -> tagToEnum# @ Q' n_a2Se;
         0# -> En.$wlvl4 n_a2Se
       }
       }
 }}}

 The point of this ticket is to consider whether it's not better to simply
 expand small number of constructors in a derived enumeration into a
 pattern match. In microbenchmark, {{{toEnum'}}} seems faster.

-- 
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/16364
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler