[GHC] #14372: CMM contains a bunch of tail-merging opportunities

GHC ghc-devs at haskell.org
Sat Jan 13 22:03:22 UTC 2018


#14372: CMM contains a bunch of tail-merging opportunities
-------------------------------------+-------------------------------------
        Reporter:  heisenbug         |                Owner:  (none)
            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 AndreasK):

 I think this can be solved in two steps:


 ----

 Change the Cmm generation for case statements:

 Currently we generate cmm of the sort
 {{{
  ...
  case 1: goto foo;
  ...

 foo:
   Register = Value
   return(Register)

 }}}

 Instead we could inline this when generating the switch.
 {{{
 #include "Cmm.h"

 test_entry()
 {
 switch [1 .. 8] R2 {
    case 1 : {R1 = 0; return();}
    ...
    case 8 : {R1 = 8; return();}
    }
 }
 }}}

 From what I've seen this hopefully won't change the generated assembly
 since GHC does this already when compiling the switch in the cases I
 looked at.

 ----


 The second step would be to change the Assembly generated to something
 like the listing in [comment:17 the comment above].

 I imagine a way to do that would be to:
 * Check if
   * Assignments are into the same register.
   * The rest of the code is the same
 * Collect all the constants, put them into a lookup table and generate
 assembly that uses the lookup table like [comment:17 above].

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14372#comment:18>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler


More information about the ghc-tickets mailing list