[GHC] #13016: SPECIALIZE INLINE doesn't necessarily inline specializations of a recursive function

[GHC]

#13016: SPECIALIZE INLINE doesn't necessarily inline specializations of a recursive
function
-------------------------------------+-------------------------------------
        Reporter:  nfrisby           |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:  Inlining
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Runtime           |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #13014, #10346    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Changes (by mpickering):

 * related:  #13014 => #13014, #10346


Comment:

 And of course, the reason that it doesn't work across modules is that
 `SpecConstr` doesn't work across modules (see #10346).

 After this analysis, I understand operationally how the program is
 optimised but I don't understand how the pragma is meant to achieve the
 unrolling if it relies on `SpecConstr` to inline a loop breaker using a
 RULE.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13016#comment:6>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler