[GHC] #14140: Better treatment for dataToTag

[GHC]

#14140: Better treatment for dataToTag
-------------------------------------+-------------------------------------
           Reporter:  simonpj        |             Owner:  (none)
               Type:  bug            |            Status:  new
           Priority:  normal         |         Milestone:
          Component:  Compiler       |           Version:  8.2.1
           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:
-------------------------------------+-------------------------------------
 In `libraries/Cabal/Cabal/Distribution/System.hs` I found
 {{{
 eqOS :: OS -> OS -> Bool
 eqOS (OtherOS s1) (OtherOS s2) = s1 == (s2 :: String)
 eqOs a b = dataToTag a == dataTag b
 }}}
 (actually it's not called `eqOS`; it's the derived equality for `OS`).

 By the time this gets to Core it looks something like this
 {{{
 eqOS a b
   = case a of
       OtherOS s1 -> case b of
                        OtherOS s2 -> eqString s1 s2
                        _ -> case dataToTag b of
                               16# -> True
                               r   -> False
       _ -> dataToTag a == dataToTag b
 }}}
 The `dataToTag` code has been duplicated; and in the `OtherOS s1` branch
 GHC can constant-fold the `dataToTag` on `a` to `16#`, the tag of
 `OtherOS`.  But GHC is no so clever for the `dataToTag b`.  We know that
 `b` is not `OtherOS`, so we know its tag is not `16#`, so the innermost
 case is entirely redundant.  But we don't spot that.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14140>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler