2 Fixed a bug, caused by the recursive use in type.h and entity.h.
3 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
4 Fixed a problem in if_while testprogram, so it now works.
7 Fixed bug in optimize(). Returned Null instead n if
8 get_optimize() == false.
11 Added some access routines in type.[hc]. Solved a recursive definition
12 of type_class struct, entity struct and type union.
15 Separated entity header into pure functional header and header
19 ... a bunch small changes not documented
22 Changed output of oo_prog_example: now uses the internal list
23 of all irgs. Had to implement some of the access routines for
24 that. Also fixed a bug in irdump. Main routines did not use
25 current_ir_graph properly.
28 Implemented some missing parts in irgopt.
29 Added field irg to entity. Only valid if entity is a method.
30 Had to resove a cyclic de[endency between irgraph.h and entity.h
32 Added fields sub and super to type_class for inheritance.
33 Added field main_irg to irprog.
36 Fixed some bugs in irgopt.c
37 Added some missing 'get_new_node' methods, so the new obstack contains
41 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
42 Solved some copy'n'paste problems. Deleted most debug-code!
43 Added comments to several files.
46 Fixed bug in _r_ constructors for all four Div operations.
47 Fixed bug in turn_into_tuple: didn't remember the block.
48 Tested transform_node.
51 Fixed some testprograms to new semantics of firm.
52 Added testprogram global_var_example.
53 Fixed a bug in irop: Size of Sel node was (entity *) instead of
54 sel_attr. Therefor array_heap_example aborted.
55 Also found a missing break in typewalk, case iro_Alloc.
56 Added a construct that unifies all things constructed by the lib
57 and needed to represent a program: irprog.ch
58 This contains a list of all types as well as a list of all ir graphs.
59 I added calls to initialization and constructors so that this is constructed
60 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
61 Removed strange block struct and constructor from ircons.c.
62 Moved routines for automatic Phi construction from irgmod to ircons.
63 Sorted functions in ircons to reflect the different interfaces.
64 Renamed dataseg to globals to be conformant with documentation.
65 Added conservation of current_ir_graph to irgopt.c.
66 Corrected bug in Makefiles: wrong path to directory ./inlude.
69 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
70 We should change the naming of the Block constructor.
71 Removed acces routines to attr. "closed" of Block.
72 Removed second implementation of new_ir_node from ircons and some comments
73 concerned with the now resolved cyclic dependency.
74 Fixed some bugs in irgopt so that the compilation goes through.
77 Changed the irnode attribut 'visit' into 'visited'.
78 Added some access functions, like get/set_irn_visit etc.
81 Renamed some access funtion in irgraph.*.
82 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
83 Added two attributes to the irgraph struct. The visited flags belongs
84 to a ir graph, not to a whole program.
87 A new optimization has been finished. The dead-node-elimination copies
88 the existing ir graph from one opstack to another. While walking the
89 graph, only the reachable nodes will be copied, so the "dead-nodes"
90 (unreachable in the controlflow) are removed.
93 Solved some problems in the 'copy_node' function. The in array was not
97 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
98 optimization, which is needed in the paramter list of 'irg_walk'.
101 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
102 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
106 Removed some files from the archive, after dependencies and usage are
108 - 'ir/common/strerror.c'
110 - 'ir/ident/xx_ident.h'
111 an empty file, only included by 'ident.c', but never used;
112 - 'ir/tv/label.c' and
114 contained a numeric counter;
115 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
116 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
117 In the old fiasco compiler, the label thing is used in the backend.
118 Search for 'babil' to see the usage.
119 Also removed the two attributes 'lab' and 'used' in the tarval struct.
122 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
123 afterwards some recursive includes, so libfirm works again.
126 Did a lot of changes, which I never commented until now (4.7.00).
128 * Added new result to Start node: Pointer to global data segment.
129 * Extended Semantics of SymConst node to represent information for the
131 * Added arithmeitc nodes (Shrs, Minus ...)
133 Rearranged the directory structure and adjusted the makefiles.
134 The directories contain:
135 ir: everything for the intermediate representation (better: src?)
136 /ir: the ir itself, and standard optimizations.
137 /tv: the target value module
138 /tr: the type and entity representation
139 /adt: abstract data types
140 /common: stuff needed by all other dirs
141 /debug: debugging Unterstuetzung
143 include: external files needen as includes
144 testprograms: examples to test the lib.
146 The makefiles generate files with extension .d that contain the dependencies
150 Added access routine to attribute link of irnode in irnode.ch.
151 Added get_negated_pnc to irnode.ch.
154 Added an iterator in iropt that calls the optimization for all nodes.
155 Copied optimize's code to optimize_in_place and removed deallocations.
156 Made set_irn_n in irnode.c public.
157 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
158 are allowded as cfgpredecessors to represent dead "incoming" control
160 Added call to optimize_in_place in mature_block. Is this useful?
163 Changed tests from comparing enums to comparing pointers. This is more
164 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
165 now get_irn_op == op_And
168 Compared equivalent_value to the one in the original fiasco.
169 Added a test in case iro_And.
170 Changed implementation of iro_Div. Added routines turn_into_tuple
171 and set_irn_op() in irgmod/irnode.
172 Changed implementation of irn_Minus (although commented out).
173 Now all features are ported or documented for this function.
176 Compared computed_value to the one in the original fiasco.
177 Added one further optimization in case iro_Proj.
178 Now all features are ported for this function.
181 Updated comments in ircons.h.
184 Changed possible attribute of SymConst from type_class to type.
185 SymConst size could be the size of a union, array, ...
188 Implemented two dumpers that dump type information and a graph with
192 Implemented a walker that walks over the type informatin that can
193 be reached from an ir node.
194 Added routine get_kind(firm_thing) in common.h, Added file common.c.
195 Added files typewalk.ch and routines to irdump.
196 Added visited flags and initialization to entity and all types.
197 Implemented some of the missing access routines in type.c, entity.c
200 Implemented constructor for ShrA.
202 Added routine that iterates the graph and verifies all nodes.
203 Added Phi_in stack to irgraph. This allows to build several
204 graphs at once with respect to the Phi building algorithm.
207 Implemented dumper for control flow graphs. Changed irr_* and dead_*
208 examples to dump cfg.
211 Implemeted explicit stack for the Phi node generation. See ircons.c.
214 added include string.h in xoprintf.c to get rid of warning:
215 implicit declaration of function `bcopy'
219 Added more cases to equivalent_node.
220 Portet gigo, identify, identify_remember.
221 The set/pset library can not handle several hash tables. At least,
222 compiling several procedures causes a segmentation fault during access
223 to the tables (oo_program_example). Need to work on this.
224 transform_node still needs to be ported.
225 some cases in equivalent_node are still missing.
227 To get it running I had to remove the call to optimize in the
228 new_Block constructor. It can only be called when the Block is
232 Introduced a walker that walks only over blocks. The asserts
233 might be too strict for general graphs, but so far it works.
234 For the walker I introduced an additional flag in the attribute
235 of Blocks, an a global block_visited variable (in irgraph.h).
236 Furhter I modularized dump_ir_node, implemented dump_ir_graph
237 with the existing walker, and added new routines to output a
239 Changed the examples to call the new dump routine.
242 Added auxiliary routines to ir_node:
243 skip_Proj, skip_Nop, is_Bad ...
244 Changed access routines to use get_irn_n instead of explicit array
245 accesses. Changed get_irn_n to call skip_Id, so that the access routines
246 never will return Id nodes.
249 Repaired Phi computation, changed new_r_Phi_in. Added test program to
250 test irregular controlflow.
253 Updated optimizations computed_value and equivalent_node to new
254 Syntax of FIRM nodes and to use access routines.
255 Added routine skip_Proj in irnode.
257 2.2.2000 Boris Boesler
258 Added files irgwalk.c and irgwalk.h. They include a simple function to
259 traverse an ir graph and execute pre and post functions to current node.
262 Added arrays with parameters and result types to the type informtion
263 of a procedure. This is needed for the code generation. Also added
264 the corresponding access routines.
265 Adapted the testprograms.
268 Found a bug / complication: the call order
269 get_value (makes Phi0, put's it into graph_arr)
270 set_value (overwrites Phi0 in graph_arr)
271 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
273 fails. Added test wether graph_arr is already set in phi_merge.
275 Added tons of comments to explain Phi generation -- the stack hack.
279 irdump now outputs the pnc_number for projs of Cmp nodes.
280 Added routine to access the pnc_string.
283 Added result "dataseg" to Start node, with all it's consequences.
284 We need this for the compiler lab to translate global variables.
285 Now they can be selected from the data segment, and the start
286 node supplies the pointer to this segment.
287 These changes are guarded by preprocessor variable UEBPRAKT in
289 Changed files: common.h, irgraph.ch, irnode.h
293 removed array_example, added array-heap_example and
294 array-stack_example and adapted makefile.
296 Entity now also allows type_method as owner. This is to model
297 the stack, Sel nodes that select a variable (array) from the
298 stack need an entity that describes what they select. This
299 entity gets the type of the enclosing procedure as owner.
302 Added future access routines in `irmode.[ch]' as comments,
303 that all attributes in the ir_mode struct can be user-defined.
304 Added access routines in `irgraph.[ch]', for all attributes.
307 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
308 Added `irvrfy.[ch]' to MAKEFILE
309 Added "inline" before access routines in `irnode.h', that no warnings
310 will appear furthermore.
311 Added three more access routines in `irmode.[ch]':
312 struct tarval *get_min_of_mode (ir_mode *mode);
313 struct tarval *get_max_of_mode (ir_mode *mode);
314 struct tarval *get_null_of_mode (ir_mode *mode);
317 Adaption of Alloc, Sel and Entity to new types:
318 Changed array_example to use type instead of type_class.
319 Changed entity to allow all types as owner instead of
320 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
321 Changed alloc to allow allocating variables of any type instead of
323 Adapted all testprograms to avoid warnings.
325 removed dtest ... from Makefile, replaced by test.
328 Added "run" to makefile, added array_example to makefile of testprograms.
329 Corrected testprogram array
330 Changed assert in irvrfy -- Conv.