2 Changed parameter of Methods in type.h and type.c from
3 class to clss. Class is a C++ keyword.
6 Changed some testprograms to accomodate Uebprakt.
9 Finished implementation of dump_all_types.
10 Added new testprogram inheritance_example.
13 Added seven access routines in type.[hc]:
14 get_class_n_member, get_class_n_subtype, get_class_n_supertype,
15 get_strct_n_member, add_strct_member, get_strct_member and set_strct_member.
16 Fixed the type_walk_2 for those seven routines.
19 Added irdump routine dump_all_types.
20 Did changes to irprog, type, typewalk to achieve this.
21 typewalk needs to walk along members, sub and supertypes, this
22 is still missing! irdump then needs to output these edges.
24 ======================================================================
25 ==================== New version: 0.2.5 ==============================
26 ======================================================================
29 Made new Version: 0.2.5
31 15.8.2000 Goetz until 3.9.00
32 Separated irnode.h and many others into two files: one public
34 Added funcionality to extract headers for lib.
37 type_dump also dumps irg->ent
38 if_example: corrected nr of locals
41 irgopt: Correct copy of self pointer in start block.
42 irgopt: Correct treatment of loops: pre function in walker
43 breaks all possible loops in Phi and Block nodes.
44 irr_loop_example: wrong Return statement
45 Now I fixed all but one error: in memory_example.
46 This error showed a conceptual problem, so I wrote
47 a slightly different algorithm. This now runs for
49 dead_node_example: cse is buggy!! merges start and the
53 looking at bug in dead_node_elimination.
55 irnode: all get_..._arr functions changed
56 ircons: new_r_Block: set matured flag.
57 ircons: new_r_SymConst: needs mode_p for linkage_ptr_info.
58 adapted call_str_example to new concepts.
61 Implemented simpler Phi construction algorithm that constructs
62 correct code for three_cfpred_example in ircons. This algorithm
63 generates more Id nodes than the original algorithm and therefore
65 Also made some changes to optimizations of Not.
68 Fixed a bug, caused by the recursive use in type.h and entity.h.
69 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
70 Fixed a problem in if_while testprogram, so it now works.
73 Fixed bug in optimize(). Returned Null instead n if
74 get_optimize() == false.
77 Added some access routines in type.[hc]. Solved a recursive definition
78 of type_class struct, entity struct and type union.
81 Separated entity header into pure functional header and header
85 ... a bunch small changes not documented
88 Changed output of oo_prog_example: now uses the internal list
89 of all irgs. Had to implement some of the access routines for
90 that. Also fixed a bug in irdump. Main routines did not use
91 current_ir_graph properly.
94 Implemented some missing parts in irgopt.
95 Added field irg to entity. Only valid if entity is a method.
96 Had to resove a cyclic de[endency between irgraph.h and entity.h
98 Added fields sub and super to type_class for inheritance.
99 Added field main_irg to irprog.
102 Fixed some bugs in irgopt.c
103 Added some missing 'get_new_node' methods, so the new obstack contains
107 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
108 Solved some copy'n'paste problems. Deleted most debug-code!
109 Added comments to several files.
112 Fixed bug in _r_ constructors for all four Div operations.
113 Fixed bug in turn_into_tuple: didn't remember the block.
114 Tested transform_node.
117 Fixed some testprograms to new semantics of firm.
118 Added testprogram global_var_example.
119 Fixed a bug in irop: Size of Sel node was (entity *) instead of
120 sel_attr. Therefor array_heap_example aborted.
121 Also found a missing break in typewalk, case iro_Alloc.
122 Added a construct that unifies all things constructed by the lib
123 and needed to represent a program: irprog.ch
124 This contains a list of all types as well as a list of all ir graphs.
125 I added calls to initialization and constructors so that this is constructed
126 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
127 Removed strange block struct and constructor from ircons.c.
128 Moved routines for automatic Phi construction from irgmod to ircons.
129 Sorted functions in ircons to reflect the different interfaces.
130 Renamed dataseg to globals to be conformant with documentation.
131 Added conservation of current_ir_graph to irgopt.c.
132 Corrected bug in Makefiles: wrong path to directory ./inlude.
135 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
136 We should change the naming of the Block constructor.
137 Removed acces routines to attr. "closed" of Block.
138 Removed second implementation of new_ir_node from ircons and some comments
139 concerned with the now resolved cyclic dependency.
140 Fixed some bugs in irgopt so that the compilation goes through.
143 Changed the irnode attribut 'visit' into 'visited'.
144 Added some access functions, like get/set_irn_visit etc.
147 Renamed some access funtion in irgraph.*.
148 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
149 Added two attributes to the irgraph struct. The visited flags belongs
150 to a ir graph, not to a whole program.
153 A new optimization has been finished. The dead-node-elimination copies
154 the existing ir graph from one opstack to another. While walking the
155 graph, only the reachable nodes will be copied, so the "dead-nodes"
156 (unreachable in the controlflow) are removed.
159 Solved some problems in the 'copy_node' function. The in array was not
160 as easy as it seamed.
163 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
164 optimization, which is needed in the paramter list of 'irg_walk'.
167 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
168 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
172 Removed some files from the archive, after dependencies and usage are
174 - 'ir/common/strerror.c'
176 - 'ir/ident/xx_ident.h'
177 an empty file, only included by 'ident.c', but never used;
178 - 'ir/tv/label.c' and
180 contained a numeric counter;
181 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
182 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
183 In the old fiasco compiler, the label thing is used in the backend.
184 Search for 'babil' to see the usage.
185 Also removed the two attributes 'lab' and 'used' in the tarval struct.
188 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
189 afterwards some recursive includes, so libfirm works again.
192 Did a lot of changes, which I never commented until now (4.7.00).
194 * Added new result to Start node: Pointer to global data segment.
195 * Extended Semantics of SymConst node to represent information for the
197 * Added arithmeitc nodes (Shrs, Minus ...)
199 Rearranged the directory structure and adjusted the makefiles.
200 The directories contain:
201 ir: everything for the intermediate representation (better: src?)
202 /ir: the ir itself, and standard optimizations.
203 /tv: the target value module
204 /tr: the type and entity representation
205 /adt: abstract data types
206 /common: stuff needed by all other dirs
207 /debug: debugging Unterstuetzung
209 include: external files needen as includes
210 testprograms: examples to test the lib.
212 The makefiles generate files with extension .d that contain the dependencies
216 Added access routine to attribute link of irnode in irnode.ch.
217 Added get_negated_pnc to irnode.ch.
220 Added an iterator in iropt that calls the optimization for all nodes.
221 Copied optimize's code to optimize_in_place and removed deallocations.
222 Made set_irn_n in irnode.c public.
223 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
224 are allowded as cfgpredecessors to represent dead "incoming" control
226 Added call to optimize_in_place in mature_block. Is this useful?
229 Changed tests from comparing enums to comparing pointers. This is more
230 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
231 now get_irn_op == op_And
234 Compared equivalent_value to the one in the original fiasco.
235 Added a test in case iro_And.
236 Changed implementation of iro_Div. Added routines turn_into_tuple
237 and set_irn_op() in irgmod/irnode.
238 Changed implementation of irn_Minus (although commented out).
239 Now all features are ported or documented for this function.
242 Compared computed_value to the one in the original fiasco.
243 Added one further optimization in case iro_Proj.
244 Now all features are ported for this function.
247 Updated comments in ircons.h.
250 Changed possible attribute of SymConst from type_class to type.
251 SymConst size could be the size of a union, array, ...
254 Implemented two dumpers that dump type information and a graph with
258 Implemented a walker that walks over the type informatin that can
259 be reached from an ir node.
260 Added routine get_kind(firm_thing) in common.h, Added file common.c.
261 Added files typewalk.ch and routines to irdump.
262 Added visited flags and initialization to entity and all types.
263 Implemented some of the missing access routines in type.c, entity.c
266 Implemented constructor for ShrA.
268 Added routine that iterates the graph and verifies all nodes.
269 Added Phi_in stack to irgraph. This allows to build several
270 graphs at once with respect to the Phi building algorithm.
273 Implemented dumper for control flow graphs. Changed irr_* and dead_*
274 examples to dump cfg.
277 Implemeted explicit stack for the Phi node generation. See ircons.c.
280 added include string.h in xoprintf.c to get rid of warning:
281 implicit declaration of function `bcopy'
285 Added more cases to equivalent_node.
286 Portet gigo, identify, identify_remember.
287 The set/pset library can not handle several hash tables. At least,
288 compiling several procedures causes a segmentation fault during access
289 to the tables (oo_program_example). Need to work on this.
290 transform_node still needs to be ported.
291 some cases in equivalent_node are still missing.
293 To get it running I had to remove the call to optimize in the
294 new_Block constructor. It can only be called when the Block is
298 Introduced a walker that walks only over blocks. The asserts
299 might be too strict for general graphs, but so far it works.
300 For the walker I introduced an additional flag in the attribute
301 of Blocks, an a global block_visited variable (in irgraph.h).
302 Furhter I modularized dump_ir_node, implemented dump_ir_graph
303 with the existing walker, and added new routines to output a
305 Changed the examples to call the new dump routine.
308 Added auxiliary routines to ir_node:
309 skip_Proj, skip_Nop, is_Bad ...
310 Changed access routines to use get_irn_n instead of explicit array
311 accesses. Changed get_irn_n to call skip_Id, so that the access routines
312 never will return Id nodes.
315 Repaired Phi computation, changed new_r_Phi_in. Added test program to
316 test irregular controlflow.
319 Updated optimizations computed_value and equivalent_node to new
320 Syntax of FIRM nodes and to use access routines.
321 Added routine skip_Proj in irnode.
323 2.2.2000 Boris Boesler
324 Added files irgwalk.c and irgwalk.h. They include a simple function to
325 traverse an ir graph and execute pre and post functions to current node.
328 Added arrays with parameters and result types to the type informtion
329 of a procedure. This is needed for the code generation. Also added
330 the corresponding access routines.
331 Adapted the testprograms.
334 Found a bug / complication: the call order
335 get_value (makes Phi0, put's it into graph_arr)
336 set_value (overwrites Phi0 in graph_arr)
337 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
339 fails. Added test wether graph_arr is already set in phi_merge.
341 Added tons of comments to explain Phi generation -- the stack hack.
345 irdump now outputs the pnc_number for projs of Cmp nodes.
346 Added routine to access the pnc_string.
349 Added result "dataseg" to Start node, with all it's consequences.
350 We need this for the compiler lab to translate global variables.
351 Now they can be selected from the data segment, and the start
352 node supplies the pointer to this segment.
353 These changes are guarded by preprocessor variable UEBPRAKT in
355 Changed files: common.h, irgraph.ch, irnode.h
359 removed array_example, added array-heap_example and
360 array-stack_example and adapted makefile.
362 Entity now also allows type_method as owner. This is to model
363 the stack, Sel nodes that select a variable (array) from the
364 stack need an entity that describes what they select. This
365 entity gets the type of the enclosing procedure as owner.
368 Added future access routines in `irmode.[ch]' as comments,
369 that all attributes in the ir_mode struct can be user-defined.
370 Added access routines in `irgraph.[ch]', for all attributes.
373 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
374 Added `irvrfy.[ch]' to MAKEFILE
375 Added "inline" before access routines in `irnode.h', that no warnings
376 will appear furthermore.
377 Added three more access routines in `irmode.[ch]':
378 struct tarval *get_min_of_mode (ir_mode *mode);
379 struct tarval *get_max_of_mode (ir_mode *mode);
380 struct tarval *get_null_of_mode (ir_mode *mode);
383 Adaption of Alloc, Sel and Entity to new types:
384 Changed array_example to use type instead of type_class.
385 Changed entity to allow all types as owner instead of
386 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
387 Changed alloc to allow allocating variables of any type instead of
389 Adapted all testprograms to avoid warnings.
391 removed dtest ... from Makefile, replaced by test.
394 Added "run" to makefile, added array_example to makefile of testprograms.
395 Corrected testprogram array
396 Changed assert in irvrfy -- Conv.