2 Added seven access routines in type.[hc]:
3 get_class_n_member, get_class_n_subtype, get_class_n_supertype,
4 get_strct_n_member, add_strct_member, get_strct_member and set_strct_member.
5 Fixed the type_walk_2 for those seven routines.
8 Added irdump routine dump_all_types.
9 Did changes to irprog, type, typewalk to achieve this.
10 typewalk needs to walk along members, sub and supertypes, this
11 is still missing! irdump then needs to output these edges.
13 ======================================================================
14 ==================== New version: 0.2.5 ==============================
15 ======================================================================
18 Made new Version: 0.2.5
20 15.8.2000 Goetz until 3.9.00
21 Separated irnode.h and many others into two files: one public
23 Added funcionality to extract headers for lib.
26 type_dump also dumps irg->ent
27 if_example: corrected nr of locals
30 irgopt: Correct copy of self pointer in start block.
31 irgopt: Correct treatment of loops: pre function in walker
32 breaks all possible loops in Phi and Block nodes.
33 irr_loop_example: wrong Return statement
34 Now I fixed all but one error: in memory_example.
35 This error showed a conceptual problem, so I wrote
36 a slightly different algorithm. This now runs for
38 dead_node_example: cse is buggy!! merges start and the
42 looking at bug in dead_node_elimination.
44 irnode: all get_..._arr functions changed
45 ircons: new_r_Block: set matured flag.
46 ircons: new_r_SymConst: needs mode_p for linkage_ptr_info.
47 adapted call_str_example to new concepts.
50 Implemented simpler Phi construction algorithm that constructs
51 correct code for three_cfpred_example in ircons. This algorithm
52 generates more Id nodes than the original algorithm and therefore
54 Also made some changes to optimizations of Not.
57 Fixed a bug, caused by the recursive use in type.h and entity.h.
58 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
59 Fixed a problem in if_while testprogram, so it now works.
62 Fixed bug in optimize(). Returned Null instead n if
63 get_optimize() == false.
66 Added some access routines in type.[hc]. Solved a recursive definition
67 of type_class struct, entity struct and type union.
70 Separated entity header into pure functional header and header
74 ... a bunch small changes not documented
77 Changed output of oo_prog_example: now uses the internal list
78 of all irgs. Had to implement some of the access routines for
79 that. Also fixed a bug in irdump. Main routines did not use
80 current_ir_graph properly.
83 Implemented some missing parts in irgopt.
84 Added field irg to entity. Only valid if entity is a method.
85 Had to resove a cyclic de[endency between irgraph.h and entity.h
87 Added fields sub and super to type_class for inheritance.
88 Added field main_irg to irprog.
91 Fixed some bugs in irgopt.c
92 Added some missing 'get_new_node' methods, so the new obstack contains
96 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
97 Solved some copy'n'paste problems. Deleted most debug-code!
98 Added comments to several files.
101 Fixed bug in _r_ constructors for all four Div operations.
102 Fixed bug in turn_into_tuple: didn't remember the block.
103 Tested transform_node.
106 Fixed some testprograms to new semantics of firm.
107 Added testprogram global_var_example.
108 Fixed a bug in irop: Size of Sel node was (entity *) instead of
109 sel_attr. Therefor array_heap_example aborted.
110 Also found a missing break in typewalk, case iro_Alloc.
111 Added a construct that unifies all things constructed by the lib
112 and needed to represent a program: irprog.ch
113 This contains a list of all types as well as a list of all ir graphs.
114 I added calls to initialization and constructors so that this is constructed
115 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
116 Removed strange block struct and constructor from ircons.c.
117 Moved routines for automatic Phi construction from irgmod to ircons.
118 Sorted functions in ircons to reflect the different interfaces.
119 Renamed dataseg to globals to be conformant with documentation.
120 Added conservation of current_ir_graph to irgopt.c.
121 Corrected bug in Makefiles: wrong path to directory ./inlude.
124 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
125 We should change the naming of the Block constructor.
126 Removed acces routines to attr. "closed" of Block.
127 Removed second implementation of new_ir_node from ircons and some comments
128 concerned with the now resolved cyclic dependency.
129 Fixed some bugs in irgopt so that the compilation goes through.
132 Changed the irnode attribut 'visit' into 'visited'.
133 Added some access functions, like get/set_irn_visit etc.
136 Renamed some access funtion in irgraph.*.
137 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
138 Added two attributes to the irgraph struct. The visited flags belongs
139 to a ir graph, not to a whole program.
142 A new optimization has been finished. The dead-node-elimination copies
143 the existing ir graph from one opstack to another. While walking the
144 graph, only the reachable nodes will be copied, so the "dead-nodes"
145 (unreachable in the controlflow) are removed.
148 Solved some problems in the 'copy_node' function. The in array was not
149 as easy as it seamed.
152 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
153 optimization, which is needed in the paramter list of 'irg_walk'.
156 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
157 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
161 Removed some files from the archive, after dependencies and usage are
163 - 'ir/common/strerror.c'
165 - 'ir/ident/xx_ident.h'
166 an empty file, only included by 'ident.c', but never used;
167 - 'ir/tv/label.c' and
169 contained a numeric counter;
170 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
171 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
172 In the old fiasco compiler, the label thing is used in the backend.
173 Search for 'babil' to see the usage.
174 Also removed the two attributes 'lab' and 'used' in the tarval struct.
177 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
178 afterwards some recursive includes, so libfirm works again.
181 Did a lot of changes, which I never commented until now (4.7.00).
183 * Added new result to Start node: Pointer to global data segment.
184 * Extended Semantics of SymConst node to represent information for the
186 * Added arithmeitc nodes (Shrs, Minus ...)
188 Rearranged the directory structure and adjusted the makefiles.
189 The directories contain:
190 ir: everything for the intermediate representation (better: src?)
191 /ir: the ir itself, and standard optimizations.
192 /tv: the target value module
193 /tr: the type and entity representation
194 /adt: abstract data types
195 /common: stuff needed by all other dirs
196 /debug: debugging Unterstuetzung
198 include: external files needen as includes
199 testprograms: examples to test the lib.
201 The makefiles generate files with extension .d that contain the dependencies
205 Added access routine to attribute link of irnode in irnode.ch.
206 Added get_negated_pnc to irnode.ch.
209 Added an iterator in iropt that calls the optimization for all nodes.
210 Copied optimize's code to optimize_in_place and removed deallocations.
211 Made set_irn_n in irnode.c public.
212 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
213 are allowded as cfgpredecessors to represent dead "incoming" control
215 Added call to optimize_in_place in mature_block. Is this useful?
218 Changed tests from comparing enums to comparing pointers. This is more
219 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
220 now get_irn_op == op_And
223 Compared equivalent_value to the one in the original fiasco.
224 Added a test in case iro_And.
225 Changed implementation of iro_Div. Added routines turn_into_tuple
226 and set_irn_op() in irgmod/irnode.
227 Changed implementation of irn_Minus (although commented out).
228 Now all features are ported or documented for this function.
231 Compared computed_value to the one in the original fiasco.
232 Added one further optimization in case iro_Proj.
233 Now all features are ported for this function.
236 Updated comments in ircons.h.
239 Changed possible attribute of SymConst from type_class to type.
240 SymConst size could be the size of a union, array, ...
243 Implemented two dumpers that dump type information and a graph with
247 Implemented a walker that walks over the type informatin that can
248 be reached from an ir node.
249 Added routine get_kind(firm_thing) in common.h, Added file common.c.
250 Added files typewalk.ch and routines to irdump.
251 Added visited flags and initialization to entity and all types.
252 Implemented some of the missing access routines in type.c, entity.c
255 Implemented constructor for ShrA.
257 Added routine that iterates the graph and verifies all nodes.
258 Added Phi_in stack to irgraph. This allows to build several
259 graphs at once with respect to the Phi building algorithm.
262 Implemented dumper for control flow graphs. Changed irr_* and dead_*
263 examples to dump cfg.
266 Implemeted explicit stack for the Phi node generation. See ircons.c.
269 added include string.h in xoprintf.c to get rid of warning:
270 implicit declaration of function `bcopy'
274 Added more cases to equivalent_node.
275 Portet gigo, identify, identify_remember.
276 The set/pset library can not handle several hash tables. At least,
277 compiling several procedures causes a segmentation fault during access
278 to the tables (oo_program_example). Need to work on this.
279 transform_node still needs to be ported.
280 some cases in equivalent_node are still missing.
282 To get it running I had to remove the call to optimize in the
283 new_Block constructor. It can only be called when the Block is
287 Introduced a walker that walks only over blocks. The asserts
288 might be too strict for general graphs, but so far it works.
289 For the walker I introduced an additional flag in the attribute
290 of Blocks, an a global block_visited variable (in irgraph.h).
291 Furhter I modularized dump_ir_node, implemented dump_ir_graph
292 with the existing walker, and added new routines to output a
294 Changed the examples to call the new dump routine.
297 Added auxiliary routines to ir_node:
298 skip_Proj, skip_Nop, is_Bad ...
299 Changed access routines to use get_irn_n instead of explicit array
300 accesses. Changed get_irn_n to call skip_Id, so that the access routines
301 never will return Id nodes.
304 Repaired Phi computation, changed new_r_Phi_in. Added test program to
305 test irregular controlflow.
308 Updated optimizations computed_value and equivalent_node to new
309 Syntax of FIRM nodes and to use access routines.
310 Added routine skip_Proj in irnode.
312 2.2.2000 Boris Boesler
313 Added files irgwalk.c and irgwalk.h. They include a simple function to
314 traverse an ir graph and execute pre and post functions to current node.
317 Added arrays with parameters and result types to the type informtion
318 of a procedure. This is needed for the code generation. Also added
319 the corresponding access routines.
320 Adapted the testprograms.
323 Found a bug / complication: the call order
324 get_value (makes Phi0, put's it into graph_arr)
325 set_value (overwrites Phi0 in graph_arr)
326 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
328 fails. Added test wether graph_arr is already set in phi_merge.
330 Added tons of comments to explain Phi generation -- the stack hack.
334 irdump now outputs the pnc_number for projs of Cmp nodes.
335 Added routine to access the pnc_string.
338 Added result "dataseg" to Start node, with all it's consequences.
339 We need this for the compiler lab to translate global variables.
340 Now they can be selected from the data segment, and the start
341 node supplies the pointer to this segment.
342 These changes are guarded by preprocessor variable UEBPRAKT in
344 Changed files: common.h, irgraph.ch, irnode.h
348 removed array_example, added array-heap_example and
349 array-stack_example and adapted makefile.
351 Entity now also allows type_method as owner. This is to model
352 the stack, Sel nodes that select a variable (array) from the
353 stack need an entity that describes what they select. This
354 entity gets the type of the enclosing procedure as owner.
357 Added future access routines in `irmode.[ch]' as comments,
358 that all attributes in the ir_mode struct can be user-defined.
359 Added access routines in `irgraph.[ch]', for all attributes.
362 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
363 Added `irvrfy.[ch]' to MAKEFILE
364 Added "inline" before access routines in `irnode.h', that no warnings
365 will appear furthermore.
366 Added three more access routines in `irmode.[ch]':
367 struct tarval *get_min_of_mode (ir_mode *mode);
368 struct tarval *get_max_of_mode (ir_mode *mode);
369 struct tarval *get_null_of_mode (ir_mode *mode);
372 Adaption of Alloc, Sel and Entity to new types:
373 Changed array_example to use type instead of type_class.
374 Changed entity to allow all types as owner instead of
375 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
376 Changed alloc to allow allocating variables of any type instead of
378 Adapted all testprograms to avoid warnings.
380 removed dtest ... from Makefile, replaced by test.
383 Added "run" to makefile, added array_example to makefile of testprograms.
384 Corrected testprogram array
385 Changed assert in irvrfy -- Conv.