[GHC] #12150: Compile time performance degradation on code that uses undefined/error with CallStacks

[GHC]

#12150: Compile time performance degradation on code that uses undefined/error with
CallStacks
-------------------------------------+-------------------------------------
        Reporter:  thomie            |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:  8.2.2
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Compile-time      |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #10844            |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Yes, this looks plausible.   (I think it's the call-stack stuff that makes
 `bool` inlined less aggressively, but I have not checked.)

 I think what we want is for the simplifier to do a "cheap and cheerful"
 bottom-ness analysis on every iteration.  See
 `CoreArity.exprBotStrictness_maybe`.

 But actually it already very nearly does exactly that.  The simplifier
 already calls `CoreArity.findRhsArity`, which in turn calls
 `CoreArity.rhsEtaEpandArity`, which calls `arityType` just like
 `exprBotStrictness_maybe`.  So we are already doing all the work (in
 `arityType`); we just aren't getting the benefit!

 I have not worked this through, but I think `findRhsArity` could return a
 boolean flag for "returns bottom" as well as an arity.  The the simplifier
 could use that to set the strictness of the `Id` if it didn't already have
 (a bottoming) strictness.

 That way `bool` would get a bottoming strictness and the case expressions
 would fall way immediately.

 In a bit more detail perhaps `SimplUtils.tryEtaExpandRhs` could return a
 new `Id` (with updated arity and perhpas strictness) as well as a new rhs.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12150#comment:16>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler