[GHC] #11736: Allow unsaturated uses of unlifted types in Core

[GHC]

#11736: Allow unsaturated uses of unlifted types in Core
-------------------------------------+-------------------------------------
        Reporter:  bgamari           |                Owner:
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:  8.2.1
       Component:  Compiler          |              Version:  8.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  GHC rejects       |  Unknown/Multiple
  valid program                      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Changes (by simonpj):

 * cc: sweirich@…, dimitris (added)


Comment:

 The issue with un-saturated type function is to do with abstraction. What
 if you had
 {{{
 (/\(a:: * -> *).  blah)  F
 }}}
 where `F` is a type function.  Well bad things, if you can decompose
 applications of `a` in `blah`.

 I think that's the only thing that can go wrong with un-saturated unlifted
 type constructors.  What if we said
 {{{
 (/\(a :: * -> *). blah)  Array#
 }}}
 Would that be OK?  Well, no: `Array# Int` is an unlifted type, not `*`.
 So it's ill-kinded.  But I think this would be fine
 {{{
 (/\(a :: TYPE 'PtrRepLifted -> TYPE 'PtrRepUnlifted). blah)  Array#
 }}}
 Inside `blah` a variable `x :: a Int` would be an unlifted type, evaluated
 call-by-value, just as it should.

 Moreover stranger things like
 {{{
 (/\(a :: Type 'PtrRepLifted -> TYPE 'VoidRep). blah) State#
 }}}
 Now inside blah a variable `x :: a Int` would be an unboxed zero-width
 void thing.  Just right.

 In short I think this is absolutely fine. We don't even need to do
 anything! (I vote for the above patch.)

 All we need is someone to come up with a proper account of runtime-rep
 polymorphism.  Kenny?

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11736#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler