2 Finished implementation of dump_all_types.
3 Added new testprogram inheritance_example.
6 Added seven access routines in type.[hc]:
7 get_class_n_member, get_class_n_subtype, get_class_n_supertype,
8 get_strct_n_member, add_strct_member, get_strct_member and set_strct_member.
9 Fixed the type_walk_2 for those seven routines.
12 Added irdump routine dump_all_types.
13 Did changes to irprog, type, typewalk to achieve this.
14 typewalk needs to walk along members, sub and supertypes, this
15 is still missing! irdump then needs to output these edges.
17 ======================================================================
18 ==================== New version: 0.2.5 ==============================
19 ======================================================================
22 Made new Version: 0.2.5
24 15.8.2000 Goetz until 3.9.00
25 Separated irnode.h and many others into two files: one public
27 Added funcionality to extract headers for lib.
30 type_dump also dumps irg->ent
31 if_example: corrected nr of locals
34 irgopt: Correct copy of self pointer in start block.
35 irgopt: Correct treatment of loops: pre function in walker
36 breaks all possible loops in Phi and Block nodes.
37 irr_loop_example: wrong Return statement
38 Now I fixed all but one error: in memory_example.
39 This error showed a conceptual problem, so I wrote
40 a slightly different algorithm. This now runs for
42 dead_node_example: cse is buggy!! merges start and the
46 looking at bug in dead_node_elimination.
48 irnode: all get_..._arr functions changed
49 ircons: new_r_Block: set matured flag.
50 ircons: new_r_SymConst: needs mode_p for linkage_ptr_info.
51 adapted call_str_example to new concepts.
54 Implemented simpler Phi construction algorithm that constructs
55 correct code for three_cfpred_example in ircons. This algorithm
56 generates more Id nodes than the original algorithm and therefore
58 Also made some changes to optimizations of Not.
61 Fixed a bug, caused by the recursive use in type.h and entity.h.
62 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
63 Fixed a problem in if_while testprogram, so it now works.
66 Fixed bug in optimize(). Returned Null instead n if
67 get_optimize() == false.
70 Added some access routines in type.[hc]. Solved a recursive definition
71 of type_class struct, entity struct and type union.
74 Separated entity header into pure functional header and header
78 ... a bunch small changes not documented
81 Changed output of oo_prog_example: now uses the internal list
82 of all irgs. Had to implement some of the access routines for
83 that. Also fixed a bug in irdump. Main routines did not use
84 current_ir_graph properly.
87 Implemented some missing parts in irgopt.
88 Added field irg to entity. Only valid if entity is a method.
89 Had to resove a cyclic de[endency between irgraph.h and entity.h
91 Added fields sub and super to type_class for inheritance.
92 Added field main_irg to irprog.
95 Fixed some bugs in irgopt.c
96 Added some missing 'get_new_node' methods, so the new obstack contains
100 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
101 Solved some copy'n'paste problems. Deleted most debug-code!
102 Added comments to several files.
105 Fixed bug in _r_ constructors for all four Div operations.
106 Fixed bug in turn_into_tuple: didn't remember the block.
107 Tested transform_node.
110 Fixed some testprograms to new semantics of firm.
111 Added testprogram global_var_example.
112 Fixed a bug in irop: Size of Sel node was (entity *) instead of
113 sel_attr. Therefor array_heap_example aborted.
114 Also found a missing break in typewalk, case iro_Alloc.
115 Added a construct that unifies all things constructed by the lib
116 and needed to represent a program: irprog.ch
117 This contains a list of all types as well as a list of all ir graphs.
118 I added calls to initialization and constructors so that this is constructed
119 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
120 Removed strange block struct and constructor from ircons.c.
121 Moved routines for automatic Phi construction from irgmod to ircons.
122 Sorted functions in ircons to reflect the different interfaces.
123 Renamed dataseg to globals to be conformant with documentation.
124 Added conservation of current_ir_graph to irgopt.c.
125 Corrected bug in Makefiles: wrong path to directory ./inlude.
128 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
129 We should change the naming of the Block constructor.
130 Removed acces routines to attr. "closed" of Block.
131 Removed second implementation of new_ir_node from ircons and some comments
132 concerned with the now resolved cyclic dependency.
133 Fixed some bugs in irgopt so that the compilation goes through.
136 Changed the irnode attribut 'visit' into 'visited'.
137 Added some access functions, like get/set_irn_visit etc.
140 Renamed some access funtion in irgraph.*.
141 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
142 Added two attributes to the irgraph struct. The visited flags belongs
143 to a ir graph, not to a whole program.
146 A new optimization has been finished. The dead-node-elimination copies
147 the existing ir graph from one opstack to another. While walking the
148 graph, only the reachable nodes will be copied, so the "dead-nodes"
149 (unreachable in the controlflow) are removed.
152 Solved some problems in the 'copy_node' function. The in array was not
153 as easy as it seamed.
156 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
157 optimization, which is needed in the paramter list of 'irg_walk'.
160 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
161 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
165 Removed some files from the archive, after dependencies and usage are
167 - 'ir/common/strerror.c'
169 - 'ir/ident/xx_ident.h'
170 an empty file, only included by 'ident.c', but never used;
171 - 'ir/tv/label.c' and
173 contained a numeric counter;
174 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
175 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
176 In the old fiasco compiler, the label thing is used in the backend.
177 Search for 'babil' to see the usage.
178 Also removed the two attributes 'lab' and 'used' in the tarval struct.
181 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
182 afterwards some recursive includes, so libfirm works again.
185 Did a lot of changes, which I never commented until now (4.7.00).
187 * Added new result to Start node: Pointer to global data segment.
188 * Extended Semantics of SymConst node to represent information for the
190 * Added arithmeitc nodes (Shrs, Minus ...)
192 Rearranged the directory structure and adjusted the makefiles.
193 The directories contain:
194 ir: everything for the intermediate representation (better: src?)
195 /ir: the ir itself, and standard optimizations.
196 /tv: the target value module
197 /tr: the type and entity representation
198 /adt: abstract data types
199 /common: stuff needed by all other dirs
200 /debug: debugging Unterstuetzung
202 include: external files needen as includes
203 testprograms: examples to test the lib.
205 The makefiles generate files with extension .d that contain the dependencies
209 Added access routine to attribute link of irnode in irnode.ch.
210 Added get_negated_pnc to irnode.ch.
213 Added an iterator in iropt that calls the optimization for all nodes.
214 Copied optimize's code to optimize_in_place and removed deallocations.
215 Made set_irn_n in irnode.c public.
216 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
217 are allowded as cfgpredecessors to represent dead "incoming" control
219 Added call to optimize_in_place in mature_block. Is this useful?
222 Changed tests from comparing enums to comparing pointers. This is more
223 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
224 now get_irn_op == op_And
227 Compared equivalent_value to the one in the original fiasco.
228 Added a test in case iro_And.
229 Changed implementation of iro_Div. Added routines turn_into_tuple
230 and set_irn_op() in irgmod/irnode.
231 Changed implementation of irn_Minus (although commented out).
232 Now all features are ported or documented for this function.
235 Compared computed_value to the one in the original fiasco.
236 Added one further optimization in case iro_Proj.
237 Now all features are ported for this function.
240 Updated comments in ircons.h.
243 Changed possible attribute of SymConst from type_class to type.
244 SymConst size could be the size of a union, array, ...
247 Implemented two dumpers that dump type information and a graph with
251 Implemented a walker that walks over the type informatin that can
252 be reached from an ir node.
253 Added routine get_kind(firm_thing) in common.h, Added file common.c.
254 Added files typewalk.ch and routines to irdump.
255 Added visited flags and initialization to entity and all types.
256 Implemented some of the missing access routines in type.c, entity.c
259 Implemented constructor for ShrA.
261 Added routine that iterates the graph and verifies all nodes.
262 Added Phi_in stack to irgraph. This allows to build several
263 graphs at once with respect to the Phi building algorithm.
266 Implemented dumper for control flow graphs. Changed irr_* and dead_*
267 examples to dump cfg.
270 Implemeted explicit stack for the Phi node generation. See ircons.c.
273 added include string.h in xoprintf.c to get rid of warning:
274 implicit declaration of function `bcopy'
278 Added more cases to equivalent_node.
279 Portet gigo, identify, identify_remember.
280 The set/pset library can not handle several hash tables. At least,
281 compiling several procedures causes a segmentation fault during access
282 to the tables (oo_program_example). Need to work on this.
283 transform_node still needs to be ported.
284 some cases in equivalent_node are still missing.
286 To get it running I had to remove the call to optimize in the
287 new_Block constructor. It can only be called when the Block is
291 Introduced a walker that walks only over blocks. The asserts
292 might be too strict for general graphs, but so far it works.
293 For the walker I introduced an additional flag in the attribute
294 of Blocks, an a global block_visited variable (in irgraph.h).
295 Furhter I modularized dump_ir_node, implemented dump_ir_graph
296 with the existing walker, and added new routines to output a
298 Changed the examples to call the new dump routine.
301 Added auxiliary routines to ir_node:
302 skip_Proj, skip_Nop, is_Bad ...
303 Changed access routines to use get_irn_n instead of explicit array
304 accesses. Changed get_irn_n to call skip_Id, so that the access routines
305 never will return Id nodes.
308 Repaired Phi computation, changed new_r_Phi_in. Added test program to
309 test irregular controlflow.
312 Updated optimizations computed_value and equivalent_node to new
313 Syntax of FIRM nodes and to use access routines.
314 Added routine skip_Proj in irnode.
316 2.2.2000 Boris Boesler
317 Added files irgwalk.c and irgwalk.h. They include a simple function to
318 traverse an ir graph and execute pre and post functions to current node.
321 Added arrays with parameters and result types to the type informtion
322 of a procedure. This is needed for the code generation. Also added
323 the corresponding access routines.
324 Adapted the testprograms.
327 Found a bug / complication: the call order
328 get_value (makes Phi0, put's it into graph_arr)
329 set_value (overwrites Phi0 in graph_arr)
330 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
332 fails. Added test wether graph_arr is already set in phi_merge.
334 Added tons of comments to explain Phi generation -- the stack hack.
338 irdump now outputs the pnc_number for projs of Cmp nodes.
339 Added routine to access the pnc_string.
342 Added result "dataseg" to Start node, with all it's consequences.
343 We need this for the compiler lab to translate global variables.
344 Now they can be selected from the data segment, and the start
345 node supplies the pointer to this segment.
346 These changes are guarded by preprocessor variable UEBPRAKT in
348 Changed files: common.h, irgraph.ch, irnode.h
352 removed array_example, added array-heap_example and
353 array-stack_example and adapted makefile.
355 Entity now also allows type_method as owner. This is to model
356 the stack, Sel nodes that select a variable (array) from the
357 stack need an entity that describes what they select. This
358 entity gets the type of the enclosing procedure as owner.
361 Added future access routines in `irmode.[ch]' as comments,
362 that all attributes in the ir_mode struct can be user-defined.
363 Added access routines in `irgraph.[ch]', for all attributes.
366 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
367 Added `irvrfy.[ch]' to MAKEFILE
368 Added "inline" before access routines in `irnode.h', that no warnings
369 will appear furthermore.
370 Added three more access routines in `irmode.[ch]':
371 struct tarval *get_min_of_mode (ir_mode *mode);
372 struct tarval *get_max_of_mode (ir_mode *mode);
373 struct tarval *get_null_of_mode (ir_mode *mode);
376 Adaption of Alloc, Sel and Entity to new types:
377 Changed array_example to use type instead of type_class.
378 Changed entity to allow all types as owner instead of
379 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
380 Changed alloc to allow allocating variables of any type instead of
382 Adapted all testprograms to avoid warnings.
384 removed dtest ... from Makefile, replaced by test.
387 Added "run" to makefile, added array_example to makefile of testprograms.
388 Corrected testprogram array
389 Changed assert in irvrfy -- Conv.