[GHC] #855: Improvements to SpecConstr

[GHC]

#855: Improvements to SpecConstr
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:  (none)
            Type:  task              |               Status:  new
        Priority:  normal            |            Milestone:  ⊥
       Component:  Compiler          |              Version:  6.4.2
      Resolution:                    |             Keywords:  SpecConstr
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Runtime           |  Unknown/Multiple
  performance bug                    |            Test Case:  N/A
      Blocked By:                    |             Blocking:  915
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by sgraf):

 Long story short, I had success with the following
 [https://github.com/sgraf812/ghc/compare/master...sgraf812:stream-fusion
 diff]. Currently, it optimizes `ex1` to this code:

 {{{
 -- RHS size: {terms: 39, types: 21, coercions: 0, joins: 3/3}
 Main.$wex1 [InlPrag=NOINLINE] :: GHC.Prim.Int# -> GHC.Prim.Int#
 [GblId, Arity=1, Caf=NoCafRefs, Str=<S,U>, Unf=OtherCon []]
 Main.$wex1
   = \ (ww_s4pl :: GHC.Prim.Int#) ->
       joinrec {
         $s$wgo_s4rk
           :: GHC.Prim.Int# -> GHC.Prim.Int# -> GHC.Prim.Int# ->
 GHC.Prim.Int#
         [LclId[JoinId(3)], Arity=3, Str=<L,A><S,U><S,U>]
         $s$wgo_s4rk _ [Occ=Dead]
                     (sc1_s4ri :: GHC.Prim.Int#)
                     (sc2_s4rh :: GHC.Prim.Int#)
           = jump $s$wgo2_s4rc sc1_s4ri sc2_s4rh;
         $s$wgo1_s4rg
           :: GHC.Prim.Int# -> GHC.Prim.Int# -> GHC.Prim.Int# ->
 GHC.Prim.Int#
         [LclId[JoinId(3)], Arity=3, Str=<S,U><S,U><S,U>]
         $s$wgo1_s4rg (sc_s4rf :: GHC.Prim.Int#)
                      (sc1_s4re :: GHC.Prim.Int#)
                      (sc2_s4rd :: GHC.Prim.Int#)
           = jump $s$wgo_s4rk sc_s4rf sc1_s4re (GHC.Prim.+# sc2_s4rd
 sc_s4rf);
         $s$wgo2_s4rc [Occ=LoopBreaker]
           :: GHC.Prim.Int# -> GHC.Prim.Int# -> GHC.Prim.Int#
         [LclId[JoinId(2)], Arity=2, Str=<S,U><S,U>, Unf=OtherCon []]
         $s$wgo2_s4rc (sc_s4rb :: GHC.Prim.Int#) (sc1_s4ra ::
 GHC.Prim.Int#)
           = case GHC.Prim.<# sc_s4rb ww_s4pl of {
               __DEFAULT -> sc1_s4ra;
               1# ->
                 jump $s$wgo1_s4rg
                   (GHC.Prim.*# sc_s4rb sc_s4rb) (GHC.Prim.+# sc_s4rb 1#)
 sc1_s4ra
             }; } in
       jump $s$wgo2_s4rc 1# 0#
 }}}

 Which is just a stone throw (or rather an inlining pass) away from the
 `goal`!

 For the examples `ex{2,3}` with increasing nesting level, things don't
 look so bright yet. Also I got some cleanup work to do tomorrow.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/855#comment:19>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler