[GHC] #14201: Implement ideas from "Compiling Pattern Matching to Good Decision Trees"

[GHC]

#14201: Implement ideas from "Compiling Pattern Matching to Good Decision Trees"
-------------------------------------+-------------------------------------
        Reporter:  AndreasK          |                Owner:  AndreasK
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 I wonder if anyone else can work with Andreas on this?  I lack bandwidth.

 Very brief thoughts:

 * Giving two or three actual concrete examples where we get better code
 would be useful.

 * Comparing ''after'' simplification is quite important.  If the
 simplifier automatically gets most of the benefit, it wouldn't be worth
 having a more complicated pattern-match compiler.

 * Do a `nofib` run with and without your patch. Is there any measurable
 difference?

 * How big is your patch?  The existing pattern match compiler is already
 quite subtle; how hard was it to implement your new scheme?

 Bottom line: to justify the extra complexity and maintenance burden, we'd
 have to be convinced that some programs run a lot faster; or that many
 programs run a bit faster.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14201#comment:12>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler