[GHC] #12984: Missed constant folding oportunities (associativity)

15 Dec 2016

      #12984: Missed constant folding oportunities (associativity)
-------------------------------------+-------------------------------------
           Reporter:  hsyl20         |             Owner:
               Type:  feature        |            Status:  new
  request                            |
           Priority:  normal         |         Milestone:
          Component:  Compiler       |           Version:  8.0.1
           Keywords:                 |  Operating System:  Unknown/Multiple
       Architecture:                 |   Type of failure:  None/Unknown
  Unknown/Multiple                   |
          Test Case:                 |        Blocked By:
           Blocking:                 |   Related Tickets:
Differential Rev(s):                 |         Wiki Page:
-------------------------------------+-------------------------------------
 Consider this simple example:
 {{{#!hs
 import System.Environment

 main :: IO ()
 main = do
    args <- getArgs
    print ((length args + 10) - 10)
 }}}

 Compiled with -O2, we get the following Core:
 {{{#!hs
 case GHC.Show.$wshowSignedInt
               0#
               (GHC.Prim.-# (GHC.Prim.+# ww2_a6Pd 10#) 10#)
               (GHC.Types.[] @ Char)
        of
 }}}

 I would expect GHC to perform constant folding there to remove the
 unnecessary operation.

 Basically, for any expression composed of Int#/Word# +/-/* primops, I
 think we should use associativity and distributivity laws to push the
 constants in the outer operation. It would allow the "scrutinee constant
 folding" optimization to be applied more often too.

 I have a patch that does this for case scrutinees, but where should I put
 it in GHC so that the optimization gets applied more generally?

--
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/12984
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler