[GHC] #7258: Compiling DynFlags is jolly slow

[GHC]

#7258: Compiling DynFlags is jolly slow
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:  simonpj
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  7.6.1
      Resolution:                    |             Keywords:  deriving-perf
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Compile-time      |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 > Maybe there is quadratic behavior there, where the number of lines of
 C-- grows proportionally with the square of the number of fields?

 But then all the C-- passes would get slower, wouldn't they?  Still it'd
 be easy to measure code size of
 * Core
 * C--
 as the number of fields increases.

 My guess is that there's an algorithmic blow-up (quadratic or worse) in
 `sink`, so that it starts to dominate when there is a very large
 `CmmProc`.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/7258#comment:59>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler