2 external to declaration of irp
3 typedef before pns and pnc_number.
6 Changed parameter of Methods in type.h and type.c from
7 class to clss. Class is a C++ keyword.
10 Changed some testprograms to accomodate Uebprakt.
13 Finished implementation of dump_all_types.
14 Added new testprogram inheritance_example.
17 Added seven access routines in type.[hc]:
18 get_class_n_member, get_class_n_subtype, get_class_n_supertype,
19 get_strct_n_member, add_strct_member, get_strct_member and set_strct_member.
20 Fixed the type_walk_2 for those seven routines.
23 Added irdump routine dump_all_types.
24 Did changes to irprog, type, typewalk to achieve this.
25 typewalk needs to walk along members, sub and supertypes, this
26 is still missing! irdump then needs to output these edges.
28 ======================================================================
29 ==================== New version: 0.2.5 ==============================
30 ======================================================================
33 Made new Version: 0.2.5
35 15.8.2000 Goetz until 3.9.00
36 Separated irnode.h and many others into two files: one public
38 Added funcionality to extract headers for lib.
41 type_dump also dumps irg->ent
42 if_example: corrected nr of locals
45 irgopt: Correct copy of self pointer in start block.
46 irgopt: Correct treatment of loops: pre function in walker
47 breaks all possible loops in Phi and Block nodes.
48 irr_loop_example: wrong Return statement
49 Now I fixed all but one error: in memory_example.
50 This error showed a conceptual problem, so I wrote
51 a slightly different algorithm. This now runs for
53 dead_node_example: cse is buggy!! merges start and the
57 looking at bug in dead_node_elimination.
59 irnode: all get_..._arr functions changed
60 ircons: new_r_Block: set matured flag.
61 ircons: new_r_SymConst: needs mode_p for linkage_ptr_info.
62 adapted call_str_example to new concepts.
65 Implemented simpler Phi construction algorithm that constructs
66 correct code for three_cfpred_example in ircons. This algorithm
67 generates more Id nodes than the original algorithm and therefore
69 Also made some changes to optimizations of Not.
72 Fixed a bug, caused by the recursive use in type.h and entity.h.
73 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
74 Fixed a problem in if_while testprogram, so it now works.
77 Fixed bug in optimize(). Returned Null instead n if
78 get_optimize() == false.
81 Added some access routines in type.[hc]. Solved a recursive definition
82 of type_class struct, entity struct and type union.
85 Separated entity header into pure functional header and header
89 ... a bunch small changes not documented
92 Changed output of oo_prog_example: now uses the internal list
93 of all irgs. Had to implement some of the access routines for
94 that. Also fixed a bug in irdump. Main routines did not use
95 current_ir_graph properly.
98 Implemented some missing parts in irgopt.
99 Added field irg to entity. Only valid if entity is a method.
100 Had to resove a cyclic de[endency between irgraph.h and entity.h
102 Added fields sub and super to type_class for inheritance.
103 Added field main_irg to irprog.
106 Fixed some bugs in irgopt.c
107 Added some missing 'get_new_node' methods, so the new obstack contains
111 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
112 Solved some copy'n'paste problems. Deleted most debug-code!
113 Added comments to several files.
116 Fixed bug in _r_ constructors for all four Div operations.
117 Fixed bug in turn_into_tuple: didn't remember the block.
118 Tested transform_node.
121 Fixed some testprograms to new semantics of firm.
122 Added testprogram global_var_example.
123 Fixed a bug in irop: Size of Sel node was (entity *) instead of
124 sel_attr. Therefor array_heap_example aborted.
125 Also found a missing break in typewalk, case iro_Alloc.
126 Added a construct that unifies all things constructed by the lib
127 and needed to represent a program: irprog.ch
128 This contains a list of all types as well as a list of all ir graphs.
129 I added calls to initialization and constructors so that this is constructed
130 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
131 Removed strange block struct and constructor from ircons.c.
132 Moved routines for automatic Phi construction from irgmod to ircons.
133 Sorted functions in ircons to reflect the different interfaces.
134 Renamed dataseg to globals to be conformant with documentation.
135 Added conservation of current_ir_graph to irgopt.c.
136 Corrected bug in Makefiles: wrong path to directory ./inlude.
139 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
140 We should change the naming of the Block constructor.
141 Removed acces routines to attr. "closed" of Block.
142 Removed second implementation of new_ir_node from ircons and some comments
143 concerned with the now resolved cyclic dependency.
144 Fixed some bugs in irgopt so that the compilation goes through.
147 Changed the irnode attribut 'visit' into 'visited'.
148 Added some access functions, like get/set_irn_visit etc.
151 Renamed some access funtion in irgraph.*.
152 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
153 Added two attributes to the irgraph struct. The visited flags belongs
154 to a ir graph, not to a whole program.
157 A new optimization has been finished. The dead-node-elimination copies
158 the existing ir graph from one opstack to another. While walking the
159 graph, only the reachable nodes will be copied, so the "dead-nodes"
160 (unreachable in the controlflow) are removed.
163 Solved some problems in the 'copy_node' function. The in array was not
164 as easy as it seamed.
167 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
168 optimization, which is needed in the paramter list of 'irg_walk'.
171 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
172 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
176 Removed some files from the archive, after dependencies and usage are
178 - 'ir/common/strerror.c'
180 - 'ir/ident/xx_ident.h'
181 an empty file, only included by 'ident.c', but never used;
182 - 'ir/tv/label.c' and
184 contained a numeric counter;
185 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
186 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
187 In the old fiasco compiler, the label thing is used in the backend.
188 Search for 'babil' to see the usage.
189 Also removed the two attributes 'lab' and 'used' in the tarval struct.
192 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
193 afterwards some recursive includes, so libfirm works again.
196 Did a lot of changes, which I never commented until now (4.7.00).
198 * Added new result to Start node: Pointer to global data segment.
199 * Extended Semantics of SymConst node to represent information for the
201 * Added arithmeitc nodes (Shrs, Minus ...)
203 Rearranged the directory structure and adjusted the makefiles.
204 The directories contain:
205 ir: everything for the intermediate representation (better: src?)
206 /ir: the ir itself, and standard optimizations.
207 /tv: the target value module
208 /tr: the type and entity representation
209 /adt: abstract data types
210 /common: stuff needed by all other dirs
211 /debug: debugging Unterstuetzung
213 include: external files needen as includes
214 testprograms: examples to test the lib.
216 The makefiles generate files with extension .d that contain the dependencies
220 Added access routine to attribute link of irnode in irnode.ch.
221 Added get_negated_pnc to irnode.ch.
224 Added an iterator in iropt that calls the optimization for all nodes.
225 Copied optimize's code to optimize_in_place and removed deallocations.
226 Made set_irn_n in irnode.c public.
227 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
228 are allowded as cfgpredecessors to represent dead "incoming" control
230 Added call to optimize_in_place in mature_block. Is this useful?
233 Changed tests from comparing enums to comparing pointers. This is more
234 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
235 now get_irn_op == op_And
238 Compared equivalent_value to the one in the original fiasco.
239 Added a test in case iro_And.
240 Changed implementation of iro_Div. Added routines turn_into_tuple
241 and set_irn_op() in irgmod/irnode.
242 Changed implementation of irn_Minus (although commented out).
243 Now all features are ported or documented for this function.
246 Compared computed_value to the one in the original fiasco.
247 Added one further optimization in case iro_Proj.
248 Now all features are ported for this function.
251 Updated comments in ircons.h.
254 Changed possible attribute of SymConst from type_class to type.
255 SymConst size could be the size of a union, array, ...
258 Implemented two dumpers that dump type information and a graph with
262 Implemented a walker that walks over the type informatin that can
263 be reached from an ir node.
264 Added routine get_kind(firm_thing) in common.h, Added file common.c.
265 Added files typewalk.ch and routines to irdump.
266 Added visited flags and initialization to entity and all types.
267 Implemented some of the missing access routines in type.c, entity.c
270 Implemented constructor for ShrA.
272 Added routine that iterates the graph and verifies all nodes.
273 Added Phi_in stack to irgraph. This allows to build several
274 graphs at once with respect to the Phi building algorithm.
277 Implemented dumper for control flow graphs. Changed irr_* and dead_*
278 examples to dump cfg.
281 Implemeted explicit stack for the Phi node generation. See ircons.c.
284 added include string.h in xoprintf.c to get rid of warning:
285 implicit declaration of function `bcopy'
289 Added more cases to equivalent_node.
290 Portet gigo, identify, identify_remember.
291 The set/pset library can not handle several hash tables. At least,
292 compiling several procedures causes a segmentation fault during access
293 to the tables (oo_program_example). Need to work on this.
294 transform_node still needs to be ported.
295 some cases in equivalent_node are still missing.
297 To get it running I had to remove the call to optimize in the
298 new_Block constructor. It can only be called when the Block is
302 Introduced a walker that walks only over blocks. The asserts
303 might be too strict for general graphs, but so far it works.
304 For the walker I introduced an additional flag in the attribute
305 of Blocks, an a global block_visited variable (in irgraph.h).
306 Furhter I modularized dump_ir_node, implemented dump_ir_graph
307 with the existing walker, and added new routines to output a
309 Changed the examples to call the new dump routine.
312 Added auxiliary routines to ir_node:
313 skip_Proj, skip_Nop, is_Bad ...
314 Changed access routines to use get_irn_n instead of explicit array
315 accesses. Changed get_irn_n to call skip_Id, so that the access routines
316 never will return Id nodes.
319 Repaired Phi computation, changed new_r_Phi_in. Added test program to
320 test irregular controlflow.
323 Updated optimizations computed_value and equivalent_node to new
324 Syntax of FIRM nodes and to use access routines.
325 Added routine skip_Proj in irnode.
327 2.2.2000 Boris Boesler
328 Added files irgwalk.c and irgwalk.h. They include a simple function to
329 traverse an ir graph and execute pre and post functions to current node.
332 Added arrays with parameters and result types to the type informtion
333 of a procedure. This is needed for the code generation. Also added
334 the corresponding access routines.
335 Adapted the testprograms.
338 Found a bug / complication: the call order
339 get_value (makes Phi0, put's it into graph_arr)
340 set_value (overwrites Phi0 in graph_arr)
341 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
343 fails. Added test wether graph_arr is already set in phi_merge.
345 Added tons of comments to explain Phi generation -- the stack hack.
349 irdump now outputs the pnc_number for projs of Cmp nodes.
350 Added routine to access the pnc_string.
353 Added result "dataseg" to Start node, with all it's consequences.
354 We need this for the compiler lab to translate global variables.
355 Now they can be selected from the data segment, and the start
356 node supplies the pointer to this segment.
357 These changes are guarded by preprocessor variable UEBPRAKT in
359 Changed files: common.h, irgraph.ch, irnode.h
363 removed array_example, added array-heap_example and
364 array-stack_example and adapted makefile.
366 Entity now also allows type_method as owner. This is to model
367 the stack, Sel nodes that select a variable (array) from the
368 stack need an entity that describes what they select. This
369 entity gets the type of the enclosing procedure as owner.
372 Added future access routines in `irmode.[ch]' as comments,
373 that all attributes in the ir_mode struct can be user-defined.
374 Added access routines in `irgraph.[ch]', for all attributes.
377 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
378 Added `irvrfy.[ch]' to MAKEFILE
379 Added "inline" before access routines in `irnode.h', that no warnings
380 will appear furthermore.
381 Added three more access routines in `irmode.[ch]':
382 struct tarval *get_min_of_mode (ir_mode *mode);
383 struct tarval *get_max_of_mode (ir_mode *mode);
384 struct tarval *get_null_of_mode (ir_mode *mode);
387 Adaption of Alloc, Sel and Entity to new types:
388 Changed array_example to use type instead of type_class.
389 Changed entity to allow all types as owner instead of
390 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
391 Changed alloc to allow allocating variables of any type instead of
393 Adapted all testprograms to avoid warnings.
395 removed dtest ... from Makefile, replaced by test.
398 Added "run" to makefile, added array_example to makefile of testprograms.
399 Corrected testprogram array
400 Changed assert in irvrfy -- Conv.