[GHC] #14610: newtype wrapping of a monadic stack kills performance

[GHC]

#14610: newtype wrapping of a monadic stack kills performance
-------------------------------------+-------------------------------------
        Reporter:  mrkkrp            |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.6.1
       Component:  Compiler          |              Version:  8.2.2
      Resolution:                    |             Keywords:  JoinPoints
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Runtime           |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #14620            |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by nomeata):

 Here is an idea. Consider a recursive function `foo` with type

 {{{
 foo :: forall t, a -> r t
 foo @t a = E[foo @t2 x `cast` co)] -- E[_] is tail call context, and the
 cast prevents join-point-hood
 }}}

 But we can transform this into a form that is allowed by now, just by
 introducing casts:

 {{{
 foo :: forall t, a -> r t
 foo @t a = go @t a refl
   where
     go :: forall t', a -> (r t ~ r t') -> r t
     go @t' a = E[go @t'2 co]
 }}}

 Note that now the return type does no longer depend on the arguments of
 `go`, so it is a join point.

 This transformation smells like a worker-wrapper transformation, but I am
 not sure.

 Anyways, I am not arguing that we should do this transformation in GHC,
 but do it in the proofs so that we can simply happily remove the
 restriction in `isValidJoinPointType`.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14610#comment:18>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler