[GHC] #14208: Performance with O0 is much better than the default or with -O2, runghc performs the best

[GHC]

#14208: Performance with O0 is much better than the default or with -O2, runghc
performs the best
-------------------------------------+-------------------------------------
        Reporter:  harendra          |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Runtime           |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by harendra):

 Alright, I guessed that ignoring the interface disables inlining so the
 problem has something to do with inlining. I looked at all the inlining
 options available and narrowed down the problem to `-fno-pre-inlining`.
 Just using `-fno-pre-inlining` brings us back to the performance that we
 are seeing with -O0. The fine manual does not say what pre-inlining is so
 I have to look at the GHC code again.

 I am confused by the fine manual. It says:

 {{{
 -fno-pre-inlining
 Default:        off
 }}}

 Does that mean pre-inlining is on by default? Can we make this clearer,
 explicit?

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14208#comment:6>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler