[libfirm] / Changes

  31.5.2000 Chris
  Removed some files from the archive, after dependences and usage are checked:
  - 'ir/common/strerror.c'
  was nowhere used;
  - 'ir/ident/xx_ident.h'
  an empty file, only included by 'ident.c', but never used;
  - 'ir/tv/label.c' and
  - 'ir/tv/label.h'
  contained a numeric counter;
  The files were included in 'tv.h' and implemented in 'tarval.c'. The two
  functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
  In the old fiasco compiler, the label thing is used in the backend.
  Search for 'babil' to see the usage.
  Also removed the two attributes 'lab' and 'used' in the tarval struct.

  26.5.2000 Chris
  Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
  afterwards some recursive includes, so libfirm works again.

*** Goetz has to complete this lines - several changes are not annotated yet! ***

  15.2.2000 Goetz
  Added access routine to attribute link of irnode in irnode.ch.
  Added get_negated_pnc to irnode.ch.

  14.2.2000 Goetz
  Added an iterator in iropt that calls the optimization for all nodes.
  Copied optimize's code to optimize_in_place and removed deallocations.
  Made set_irn_n in irnode.c public.
  Added Bad as possible cfg predecessor in is_cfg_node in irnode.  Bads
  are allowded as cfgpredecessors to represent dead "incoming" control
  flow.
  Added call to optimize_in_place in mature_block. Is this useful?

  10.2.2000 Goetz
  Changed tests from comparing enums to comparing pointers. This is more
  efficient and reads better.  e.g., instead get_irn_opcode == irm_And
  now get_irn_op == op_And

  10.2.2000 Goetz
  Compared equivalent_value to the one in the original fiasco.
  Added a test in case iro_And.
  Changed implementation of iro_Div. Added routines turn_into_tuple
    and set_irn_op() in irgmod/irnode.
  Changed implementation of irn_Minus (although commented out).
  Now all features are ported or documented for this function.

  10.2.2000 Goetz
  Compared computed_value to the one in the original fiasco.
  Added one further optimization in case iro_Proj.
  Now all features are ported for this function.

  9.2.2000 Goetz
  Updated comments in ircons.h.

  9.2.2000 Goetz
  Changed possible attribute of SymConst from type_class to type.
  SymConst size could be the size of a union, array, ...

  9.2.2000 Goetz
  Implemented two dumpers that dump type information and a graph with
  type information.

  9.2.2000 Goetz
  Implemented a walker that walks over the type informatin that can
  be reached from an ir node.
  Added routine get_kind(firm_thing) in common.h, Added file common.c.
  Added files typewalk.ch and routines to irdump.
  Added visited flags and initialization to entity and all types.
  Implemented some of the missing access routines in type.c, entity.c

  9.2.2000 Goetz
  Implemented constructor for ShrA.
  Added type struct.
  Added routine that iterates the graph and verifies all nodes.
  Added Phi_in stack to irgraph.  This allows to build several
    graphs at once with respect to the Phi building algorithm.

  9.2.2000 Goetz
  Implemented dumper for control flow graphs.  Changed irr_* and dead_*
  examples to dump cfg.

  8.2.2000 Goetz
  Implemeted explicit stack for the Phi node generation.  See ircons.c.

  8.2.2000 Goetz
  added include string.h in xoprintf.c to get rid of warning:
  implicit declaration of function `bcopy'

  7.2.2000 Goetz
  Worked on iropt.c:
  Added more cases to equivalent_node.
  Portet gigo, identify, identify_remember.
  The set/pset library can not handle several hash tables.  At least,
  compiling several procedures causes a segmentation fault during access
  to the tables (oo_program_example).  Need to work on this.
  transform_node still needs to be ported.
  some cases in equivalent_node are still missing.

  To get it running I had to remove the call to optimize in the
  new_Block constructor. It can only be called when the Block is
  mature.

  7.2.2000 Goetz
  Introduced a walker that walks only over blocks.  The asserts
  might be too strict for general graphs, but so far it works.
  For the walker I introduced an additional flag in the attribute
  of Blocks, an a global block_visited variable (in irgraph.h).
  Furhter I modularized dump_ir_node, implemented dump_ir_graph
  with the existing walker, and added new routines to output a
  basic block graph.
  Changed the examples to call the new dump routine.

  7.2.2000 Goetz
  Added auxiliary routines to ir_node:
  skip_Proj, skip_Nop, is_Bad ...
  Changed access routines to use get_irn_n instead of explicit array
  accesses. Changed get_irn_n to call skip_Id, so that the access routines
  never will return Id nodes.

  3.2.2000 Goetz
  Repaired Phi computation, changed new_r_Phi_in.  Added test program to
  test irregular controlflow.

  2.2.2000 Goetz
  Updated optimizations computed_value and equivalent_node to new
  Syntax of FIRM nodes and to use access routines.
  Added routine skip_Proj in irnode.

  2.2.2000 Boris Boesler
  Added files irgwalk.c and irgwalk.h. They include a simple function to
  traverse an ir graph and execute pre and post functions to current node.

  1.2.2000 Goetz
  Added arrays with parameters and result types to the type informtion
  of a procedure.  This is needed for the code generation.  Also added
  the corresponding access routines.
  Adapted the testprograms.

  1.2.2000 Goetz
  Found a bug / complication:  the call order
    get_value   (makes Phi0, put's it into graph_arr)
    set_value   (overwrites Phi0 in graph_arr)
    mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
                  the proper value.)
  fails.  Added test wether graph_arr is already set in phi_merge.

  Added tons of comments to explain Phi generation -- the stack hack.


  27.1.2000 Goetz
  irdump now outputs the pnc_number for projs of Cmp nodes.
  Added routine to access the pnc_string.

  27.1.2000 Goetz
  Added result "dataseg" to Start node, with all it's consequences.
  We need this for the compiler lab to translate global variables.
  Now they can be selected from the data segment, and the start
  node supplies the pointer to this segment.
  These changes are guarded by preprocessor variable UEBPRAKT in
  common.h
  Changed files: common.h, irgraph.ch, irnode.h

  26.1.2000 Goetz
  Testprograms:
  removed array_example, added array-heap_example and
  array-stack_example and adapted makefile.
  ir_lib:
  Entity now also allows type_method as owner. This is to model
  the stack, Sel nodes that select a variable (array) from the
  stack need an entity that describes what they select.  This
  entity gets the type of the enclosing procedure as owner.

  24.1.2000 Chris
  Added future access routines in `irmode.[ch]' as comments,
    that all attributes in the ir_mode struct can be user-defined.
  Added access routines in `irgraph.[ch]', for all attributes.

  22.1.2000 Chris
  Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
  Added `irvrfy.[ch]' to MAKEFILE
  Added "inline" before access routines in `irnode.h', that no warnings
    will appear furthermore.
  Added three more access routines in `irmode.[ch]':
        struct tarval *get_min_of_mode (ir_mode *mode);
        struct tarval *get_max_of_mode (ir_mode *mode);
        struct tarval *get_null_of_mode (ir_mode *mode);

  21.1.2000 Goetz
  Adaption of Alloc, Sel and Entity to new types:
    Changed array_example to use type instead of type_class.
    Changed entity to allow all types as owner instead of
      only type_class.  Added routine "assert_legal_owner_of_ent(type* type)".
    Changed alloc to allow allocating variables of any type instead of
      only type_class.
    Adapted all testprograms to avoid warnings.

  removed dtest ... from Makefile, replaced by test.

  20.1.2000 Goetz
  Added "run" to makefile, added array_example to makefile of testprograms.
  Corrected testprogram array
  Changed assert in irvrfy -- Conv.
  Started Change log.
  Started TODO log.