Modifications of GHC for HasFuse
                   ================================

                   (c) David Sabel  September 2003

This file contains a list of modfications of the ghc-5.04.3 source code
which have been done for for HasFuse by David Sabel 


NOTICE: 

1) Modified or added lines in the source code are marked with [DS], so
   you can easy search for those lines.

2) There's a HFusMod.pdf file with the same content as this file, but
   much better for printing.
	


************************************************************************
**
** file: ghc/compiler/main/DriverFlags.hs
**
** comments:
**    - I added a new flag --hasfuse, which shows some information about
**      HasFuse
**
**    - I added a new flag -fstrictness, which turns strictness analysis
**      on, because strictness-analysis is off by default at
**      optimisation level 0 and 1 (see comments about DriverState.hs)
**
************************************************************************


************************************************************************
**
** file: ghc/compiler/main/DriverState.hs
**
** comments:
**    - changes in hsc_minusO_flags:
**        + ignore interface-pragmas
**        + omit interface-pragmas
**        + don't do the foldr-build
**        + turn off cpr
**
**    - changes in hsc_minusO2_flags:
**        There's a real definition now, because we turn foldr-build on
**        and cpr not off.
**        The rest of the flags are same as -O1 flags
**
**    - big changes in buildCoreToDo for -O1 and -O2:
**
**      + strictness analysis is now off by default at level 0 and 1, 
**        but it can be switched on by the flag -fstrictness 
**        (Worker Wrapper is then not on!)
**
**      + you can't switch off strictness analysis at level 2
**        (the flag -fno-strictness has no effect)
**
**      + level 1:
**         Do only "Floating In" and run the Simplifier
**      + level 2:
**         Do every optimisation as before, that aren't tested for 
**         safeness. 
**
**      + The unsafe transformations FloatOut and CSE aren't performed
**        in any optimisation level!
**        The liberate case transformation is also never performed,
**        because we have nearly no documentation about it.
**
************************************************************************



************************************************************************
**
** file: ghc/driver/ghc-usage.txt
**
** comments:
**    - I added information about HasFuse and the --hasfuse flag
**
**    - I added information for use of optimisation level 2, which isn't
**      proven as 'safe'
**
************************************************************************



************************************************************************
**
** file: ghc/mk/version.mk
**
** comments: 
**    - I changed the Projectname to HasFuse
**
************************************************************************



************************************************************************
**
** file: ghc/compiler/coreSyn/CoreUtils.lhs
**
**
** comments:
**    - I've made some changes in the definition of exprIsCheap:
**
**      + An expression is now cheap, if it's a variable, an unapplied 
**        primitive operator with positive arity, a literal, a 
**        constructor or a constructor application where the arguments
**        are cheap.
**
**      + I also allow things about types to be cheap, because FUNDIO 
**        has no types.
**
**      + Primitive operators are treated like abstractions, so
**        primitive operators with arity 0 aren't cheap
**
**    - I've made some changes in the definition of exprIsTrivial:
**
**      +  I don't allow to duplicate primitive operators or 
**         foreign-call Ids with arity 0, so I changed exprIsTrivial for
**         the  Var case.
**
**    - I replaced the definition of exprEtaExpandArity by exprArity,
**      because the eta-expansion performed by the GHC is not safe in
**      the sense of FUNDIO.
**
**    - I've made some changes in the definition of exprIsValue:
**
**      + Partial applications are no longer treated as values in general.
**        I only allow applications to constructors with to few 
**        arguments as values.
**
**      + exprOkForSpeculation is no longer used in exprIsValue, 
**        it's replaced by False. It would be nice to turn off 
**        exprOkForSpeculation fully. I tried that with the definition 
**           exprOkForSpeculation = False,
**        but then the compiler fails while building the libraries
**        (Data/Array/Base.hs).
**
************************************************************************




************************************************************************
**
** file: ghc/compiler/simplCore/OccurAnal.lhs
**
** line(s): 621
**
** comment:
**    - I've made some changes in occurAnal for the Lam case, so 
**      that InsideLam is now also True for one-shot-lambdas.
**
************************************************************************



************************************************************************
**
** file: ghc/compiler/simplCore/SimplUtils.lhs
**
** line(s): 621
**
** comment: 
**    - I changed mkCase1:
**      We don't check for exprOkForSpeculation, because we don't know 
**      if case elimination is correct in this case.
**
************************************************************************



************************************************************************
**
** file: ghc/compiler/main/Main.hs
**
** line(s): 213
**
** comment: 
**    - We print a new message for the modified version (HasFuse), each 
**      time GHC is running.
**
**    - I changed the verbosity information for HasFuse.
**
************************************************************************



************************************************************************
**
** file: ghc/compiler/simplCore/SimplMonad.lhs
**
** line(s):
**
** comment:
**    - I turned off RULES completely, by changing the definition of 
**      activeRule, so that the result is always Nothing.
**
************************************************************************



************************************************************************
**
** file: ghc/compiler/utils/Panic.lhs
**
** comment:
**    - I changed the panic-message in showGhcException (Panic s),
**      bugs should be reported to the HasFuse webpage.
**
************************************************************************



************************************************************************
**
** file: ghc/compiler/main/SysTool.lhs
**
** comments:
**    - I added a new function showHasFuseMess, that's used for printing
**      the HasFuse message, which comes with flag --hasfuse.
**      Therefore I added the global var v_Path_hasfusemess with
**      initialisation
**
************************************************************************



************************************************************************
** file: ghc/driver/hfmess.txt
**
** comments:
**    - This file is new, it contains the message that comes with 
**      --hasfuse
**
************************************************************************