[GHC] #5245: Cmm optimizer: propagate constants across basic block boundaries

[GHC]

#5245: Cmm optimizer: propagate constants across basic block boundaries
------------------------------+---------------------------------------------
  Reporter:  tibbe            |          Owner:                  
      Type:  feature request  |         Status:  closed          
  Priority:  normal           |      Milestone:                  
 Component:  Compiler         |        Version:  7.0.3           
Resolution:  fixed            |       Keywords:                  
  Testcase:                   |      Blockedby:                  
Difficulty:                   |             Os:  Unknown/Multiple
  Blocking:                   |   Architecture:  Unknown/Multiple
   Failure:  None/Unknown     |  
------------------------------+---------------------------------------------
Changes (by ezyang):

  * status:  new => closed
  * resolution:  => fixed


Comment:

 With new codegen assignment rewriting and machine op folding, this
 optimization should be happening for the new codegen. Since simonpj is not
 interested in fixing this in the old path, I'm marking this as fixed. If
 there are specific examples where this is still failing, please report
 them!

 {{{
 ezyang at javelin:~$ Dev/ghc-build-master-stage2-default/inplace/bin/ghc-
 stage2 -c test.cmm  -ddump-opt-cmm -O -fconvert-to-zipper-and-back

 ==================== Optimised Cmm ====================
 fn()    { []
         }
     cd: R1 = 2 :: W64;
         jump (I32[Sp + 0]) ();
 }
 }}}

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5245#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler