[GHC] #14254: The Binary instance for TypeRep smells a bit expensive

[GHC]

#14254: The Binary instance for TypeRep smells a bit expensive
-------------------------------------+-------------------------------------
        Reporter:  dfeuer            |                Owner:  (none)
            Type:  bug               |               Status:  patch
        Priority:  normal            |            Milestone:  8.4.1
       Component:  Compiler          |              Version:  8.2.1
      Resolution:                    |             Keywords:  Typeable
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Compile-time      |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):  Phab:D3998
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by bgamari):

 I think you need to define what your cost model is; it's not clear that
 anyone expects `dynApply` to be as cheap as a static application. The same
 goes for serialisation; it's going to be expensive even if you cache the
 kind. This is why packages like `distributed-process` go to some lengths
 to avoid serialising unnecessarily. On the other hand, packages like
 `vault` really don't care how expensive kind computation is, so optimizing
 for this case may be counterproductive from their perspective.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14254#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler