[GHC] #14068: Loopification using join points

[GHC]

#14068: Loopification using join points
-------------------------------------+-------------------------------------
        Reporter:  nomeata           |                Owner:  nomeata
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:  JoinPoints
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #13966 #14067     |  Differential Rev(s):  Phab:D3811
  #14827                             |
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by nomeata):

 > So why is this worse than the previous version?

 Because if `drop` itself is in an inner loop, something like `map (\n ->
 drop n xs) [0..1000]`, then we have to allocate a box for the the `I#` to
 wrap the argument.

 Previously, `drop` would have a specialization rule, and the calls to
 `drop` would be replaced by a call to `$sdrop`.

 _But_: Maybe we should simply not worry: For x2n1, allocation goes up by
 80%, but the instruction count stays the same (+0.05%). I guess that the
 extra `I#` constructor needed to allocate is consumed pretty quickly, and
 never GC’ed…

 _But again_: This line of reasoning doesn’t cut it for other benchmarks.
 For example `treejoin`’s allocation goes up by 27%, and the instruction
 count by 72%(!). With `-f-no-spec-constr`, there is no difference in
 either number.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14068#comment:36>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler