[GHC] #15056: Wrappers inlined too late

[GHC]

#15056: Wrappers inlined too late
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.6.1
       Component:  Compiler          |              Version:  8.2.2
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Another example comes from `nofib/real/veritas`. In `Edlib.hs` we have
 {{{
 (...) :: Xio_fn -> Xio_fn -> Xin -> Xio
 (...) f g xin = f xin \\\ g
 }}}
 and  in `X_interface` we get
 {{{
 x_send_argL argL = x_send x_multi_arg_esc_seq      ...
                    x_send [toEnum (length argL)]   ...
                    app ( map x_send_arg argL )     ...
                    x_send x_end_arg_esc_seq
 }}}
 It turns out that `...` gets w/w'd (because it is CPR'd in fact), and that
 means it doesn't inline until phase 0.  If we inline in phase 2 we get
 lots
 of fold/build fusion, which leads to a whopping 25% improvement in the
 allocation of the veritas benchmark.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15056#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler