[GHC] #9701: GADTs not specialized properly

[GHC]

#9701: GADTs not specialized properly
-------------------------------------+-------------------------------------
              Reporter:  dfeuer      |            Owner:
                  Type:  bug         |           Status:  new
              Priority:  normal      |        Milestone:
             Component:  Compiler    |          Version:  7.9
            Resolution:              |         Keywords:
      Operating System:              |     Architecture:  Unknown/Multiple
  Unknown/Multiple                   |       Difficulty:  Unknown
       Type of failure:  Runtime     |       Blocked By:
  performance bug                    |  Related Tickets:
             Test Case:              |
              Blocking:              |
Differential Revisions:              |
-------------------------------------+-------------------------------------

Comment (by dfeuer):

 The reason I set about this whole line of thinking was the fact that `Set`
 cannot implement an efficient `Data.Foldable.elem` method. Wrapping a
 `Set` in an `Ord`-carrying GADT would enable such a thing. Unfortunately,
 if the GADT blocks `Ord` specialization, everything will be rather slow:
 comparison operators will always be looked up in the dictionary the
 wrapper points to, and there will be no opportunity for inlining, etc. My
 position regarding Incoherent Instances (and DysFunctional Dependencies)
 is that if your program breaks when the compiler hands you a different
 dictionary than you expected, you get to keep both pieces—humans invented
 nasal demons for a reason. I'm much more open to arguments against this
 idea based on implementation complexity.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9701#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler