[GHC] #14137: Do more inlining into non-recursive join points

[GHC]

#14137: Do more inlining into non-recursive join points
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:  nomeata
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.1
      Resolution:                    |             Keywords:  JoinPoints
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 nomeata):

 >  I think `isJoinId` would be fine, and much faster.

 Isn’t it the simplifier that turns a potential join point into one? So if
 I use `isJoinId`, won’t this delay the effect of the patch to the first
 run off the occurrence analyser after a simplifier run, instead of doing
 it right in the first occurrence analyzer run? (`tagNonRecBinder` only
 modifies the `idOccInfo`).

 >  Not quite: ds5 won't be a join point; it'll disappear entirely by being
 inlined.

 What about the many occurrences of `ds5` in `safe`? Don’t they prevent
 inlining, due to code duplication?

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14137#comment:6>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler