occ-960117

[acken@informatik.uni-kl.de]

Looks like Juergen forgot to send out a mail that the 960117 sources
can be found on the cognac site.  So I'm sending out this note in his 
place.  

What's new?
* had to rewrite the size/alignment calculation for imported types 
  again, fields of extended records still got wrong offsets; hope I
  finally got it right 
* initial changes to allow the front-end to keep translations of
  multiple modules in memory; this is intended for a global
  optimization pass that looks at the GSA code of all modules of the
  program simultaneously 
* Juergen is still changing constant propagation and algebraic
  transformations (note that const prop is probably still broken)
* some minor bufixes

-- mva


Tue Jan 16 16:25:55 1996  Juergen Zimmermann  <jnzimmer@informatik.uni-kl.de>

	* optimizers/ValueNumbering.Mod: *** empty log message ***

	* optimizers/AlgebraicTransformation.Mod:
	The module is still not finished.

	Introduced `ArithmTransform', a procedure for all those special
	arithmetical cases which were not catched yet.

	* optimizers/ConstPropagation.Mod:
	Also have a look on `mergeCond' the same way as we do on `mergeIf'/`mergeCase'.
	This way, some more `gates' become eliminated.

	Rewrote parts of the code and replaced them by smaller, equivalent ones.

Mon Jan 15 20:18:59 1996  Michael van Acken  <acken@informatik.uni-kl.de>

	* frontend/Data.Mod, frontend/Make.Mod, frontend/ParseDecl.Mod, frontend/ParseExpr.Mod, frontend/ParseStatm.Mod, frontend/SymbolTable.Mod, frontend/WriteGSA.Mod:
	Splitted `Object.mnolev' into `Object.moduleId' and `Object.level'.

	* backend/ansi-c/StdTypes.Mod: Fixed some comments.

	* tools/Browser.Mod (WriteSet):
	Initial "{" of set constants was missing.

	* frontend/SymbolTable.Mod (ReadSymbolFile):
	Calculation of size and alignment of imported types
	was still broken for some extended records.  Changed import to do the
	calculation when all types are in memory, and not as part of the
	reading process.

	* frontend/Error.Mod:
	Extended module to handle error messages for multiple source files.
	Adapted error suppression to the one used by o2c.

Mon Jan 15 14:50:21 1996  Juergen Zimmermann  <jnzimmer@informatik.uni-kl.de>

	* optimizers/ConstPropagation.Mod:
	Reworked parts of `ReplaceByConstants' completely, so that much redundant
	code is ommited.

Mon Jan 15 10:30:25 1996  Michael van Acken  <acken@informatik.uni-kl.de>

	* frontend/ErrorInstr.Mod: Initial revision

	* backend/ansi-c/StdTypes.Mod: Exported min/max constants.

Sun Jan 14 20:01:41 1996  Michael van Acken  <acken@informatik.uni-kl.de>

	* frontend/ParseExpr.Mod (Designator):
	Don't erase reference to open array length until the
	designator has reached the end of the open dimensions.

	* frontend/ParseStatm.Mod (ReclaimFromMem):
	Don't reclaim values that the called function
	returns in memory anyway.
	(RemoveCollectedOpnds): New.  Don't collect values into mem that the
	following call will force into mem anyway.

	(PassParameter): Added location attributes for values passed to formal
	parameters by a `call' instruction.  The formal parameter objects are
	used for the location's `var' field.
	(RegisterSideEffects): Fixed faulty location flags.

	* frontend/Data.Mod (GetModule):
	New.  Takes a procedure or module object, returns the
	top level module it is part of.

Thu Jan 11 15:13:49 1996  Juergen Zimmermann  <jnzimmer@informatik.uni-kl.de>

	* optimizers/Worklist.Mod: Removed unused stuff.

	* optimizers/AlgebraicTransformation.Mod: Removed `HandleGuard'.
	Rewrote `BooleanResult' in order to handle `NOT' instructions better.

	* optimizers/ConstPropagation.Mod:
	Rewrote `ReplaceByConstant' completely. Now moves instructions correctly
	and inspects regions only once. Not reached regions are deleted in an extra
	phase.

Tue Jan  9 11:23:14 1996  Juergen Zimmermann  <jnzimmer@informatik.uni-kl.de>

	* optimizers/ConstPropagation.Mod:
	Intermediate version, some fixes by Michael van Acken.