[GHC] #11735: Optimize coercionKind

[GHC]

#11735: Optimize coercionKind
-------------------------------------+-------------------------------------
        Reporter:  goldfire          |                Owner:  (none)
            Type:  task              |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  7.10.3
      Resolution:                    |             Keywords:
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 tdammers):

 Re-running my profiling job on a pristine machine without significant
 background noise gives me this:

 {{{
 --- ./cases/Grammar.hs ---
 d74b37d565: 16.21
 8ac966971e: 12.87
 4eb140f564: 12.38
 73a99750e1: 20.00
 4572849929: 19.67
 452dee3ff4: 256.07
 --- ./cases/test_rules2.hs ---
 d74b37d565: 1.12
 8ac966971e: 1.10
 4eb140f564: 1.10
 73a99750e1: 1.10
 4572849929: 1.07
 452dee3ff4: 1.03
 }}}

 Again, with meanings:
 {{{
 d74b37d565: simplCore improvements
 8ac966971e: With NthCo Role caching
 4eb140f564: Un-refactored
 73a99750e1: Added some SCC's
 4572849929: Simon's patch applied
 452dee3ff4: baseline, GHC `master` before branching off
 }}}

 Which mostly confirms the earlier conclusions. The `test_rules2.hs`
 example seems to get worse as we proceed through the changes though.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11735#comment:54>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler