Re: [GHC] #8996: mark more things as const in C codegen

16 Apr 2014

      #8996: mark more things as const in C codegen
---------------------------------------+-----------------------------------
        Reporter:  slyfox              |            Owner:
            Type:  feature request     |           Status:  new
        Priority:  normal              |        Milestone:
       Component:  Compiler (CodeGen)  |          Version:  7.8.2
      Resolution:                      |         Keywords:
Operating System:  Unknown/Multiple    |     Architecture:
 Type of failure:  None/Unknown        |  Unknown/Multiple
       Test Case:                      |       Difficulty:  Unknown
        Blocking:                      |       Blocked By:
                                       |  Related Tickets:
---------------------------------------+-----------------------------------
Changes (by slyfox):

 * difficulty:  Easy (less than 1 hour) => Unknown

Comment:

 This patch attempts to mark everything ghc knows as RODATA,
 what raises a couple of questions:

 Patch can be tested as './configure --enable-unregisterised' in amd64.

 Example of build failure:
 {{{
   HC [stage 1] libraries/ghc-prim/dist-install/build/GHC/Types.o

 In file included from /tmp/ghc26762_0/ghc26762_1.hc:3:0:

 /tmp/ghc26762_0/ghc26762_1.hc:295:6:
      error: conflicting types for 'ghczmprim_GHCziTypes_False_closure'
      CEI_(ghczmprim_GHCziTypes_False_closure);
           ^

 /home/slyfox/dev/git/ghc/includes/Stg.h:213:52:
      note: in definition of macro 'CEI_'
      #define CEI_(X)         extern const StgWordArray (X)
 GNU_ATTRIBUTE(aligned (8))
                                                         ^

 /tmp/ghc26762_0/ghc26762_1.hc:27:9:
      note: previous definition of 'ghczmprim_GHCziTypes_False_closure' was
 here
      StgWord ghczmprim_GHCziTypes_False_closure[] = {
              ^
 }}}

 GHC thinks that extern points to rodata, but actually place or to .data
 (rwdata).
 I think it's due to '''rts/StgMiscClosures.cmm''':
 {{{
 /* put these in the *data* section, since the garbage collector relies
  * on the fact that static closures live in the data section.
  */

 #if !(defined(COMPILING_WINDOWS_DLL))
 section "data" {
  stg_CHARLIKE_closure:
     CHARLIKE_HDR(0)
     CHARLIKE_HDR(1)
     CHARLIKE_HDR(2)
     CHARLIKE_HDR(3)
 ...
 }}}

 Is it an absolute need to use mutable data section for it?
 Could GC be tuned to scan for certain rodata block?

 Or if it's not an option,

 fix ghc to know that those closure externs
 are really in "data" section.

 Thanks!

--
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8996#comment:3
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler