[GHC] #4344: Better toRational for Float and Double

[GHC]

#4344: Better toRational for Float and Double
----------------------------------------+-----------------------------------
    Reporter:  daniel.is.fischer        |        Owner:                         
        Type:  bug                      |       Status:  new                    
    Priority:  normal                   |    Milestone:                         
   Component:  Compiler                 |      Version:  6.12.3                 
    Keywords:  toRational, performance  |     Testcase:                         
   Blockedby:                           |   Difficulty:                         
          Os:  Unknown/Multiple         |     Blocking:                         
Architecture:  Unknown/Multiple         |      Failure:  Runtime performance bug
----------------------------------------+-----------------------------------

Comment(by daniel.is.fischer):

 Well, I would replace the body of `toRational` in the `Real` instances of
 `Float` and `Double` with more or less the body of `toRat3`.

 The bang on `d` makes no difference with optimisations and little
 difference without, so that'll go. If testing also for `e == 0` makes a
 difference at all, it's minuscule (less than 1%), so I'd leave that out.

 I would like to test whether something can be gained by determining the
 number of trailing 0-bits of `m` to avoid all `gcd`s before I propose a
 specific patch.

 Since GHC.Float is GHC-internal and maintainer is cvs-ghc and not
 libraries, should the proposal go through the process nevertheless or
 would I just attach the patch here and you validate and push directly?

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/4344#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler