[GHC] #1600: Optimisation: CPR the results of IO

[GHC]

#1600: Optimisation: CPR the results of IO
-------------------------------------+-------------------------------------
        Reporter:  simonmar          |                Owner:  (none)
            Type:  task              |               Status:  new
        Priority:  lowest            |            Milestone:
       Component:  Compiler          |              Version:  6.6.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Runtime           |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #8598             |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 > The main difference seems to come from ​this change...

 Interesting!  There are two changes in that commit:

 * Making data con wrappers have `ug_unsat_ok = unSaturatedOk`.  I buy
 this. Can you try the effect of this change alone? It's entirely unrelated
 to nested-CPR.  It would be good to isolate that perf bump from the perf
 changes due to nested CPR.

 * Making data con wrapper have `ug_boring_ok = boringCxtOk`.  I'm not sure
 I buy this in full.  In `Note [Inline data constructor wrappers
 aggresively]` there's a claim that we get better nested-CPR info.  That's
 true if we have a polymorphic bang argument (e.g. `data T a = MkT !a`) but
 for a monomorphic one like `data S = MkS !Int` we'll unpack it.

   So I think the benefit comes only for polymorphic bangs.  And there is a
 cost to aggressive inlining, so we don't want to do it gratuitously.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/1600#comment:65>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler