[GHC] #947: ghc -O space leak: CSE between different CAFs

[GHC]

#947: ghc -O space leak: CSE between different CAFs
---------------------------+------------------------------------------------
  Reporter:  int-e@…       |          Owner:         
      Type:  bug           |         Status:  new    
  Priority:  normal        |      Milestone:  _|_    
 Component:  Compiler      |        Version:  6.5    
Resolution:                |       Keywords:         
        Os:  Linux         |   Architecture:  x86    
   Failure:  None/Unknown  |     Difficulty:  Unknown
  Testcase:                |      Blockedby:         
  Blocking:                |        Related:         
---------------------------+------------------------------------------------

Comment(by int-e):

 Replying to [comment:9 simonpj]:
 > I have one other very simple suggestion: '''do no CSE for top-level
 CAFs'''.

 Sounds like a great idea to me. An added benefit is that it would lend
 more safety to constructing top-level mutable variables (like
 {{{IORef}}}s). The loss of sharing should be negligible for most purposes,
 but I can think of a few cases where giving names to shared parts would be
 awkward:

  * repeated string constants when doing text processing
  * repeated numerical values, in particular large constants containing
 repetitions (the worst of this can probably be avoided by just doing CSE
 before floating stuff out so that {{{[0,1,0,1]}}} ends up with the zeros
 and ones shared).

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/947#comment:10>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler