2 Added irdump routine dump_all_types.
3 Did changes to irprog, type, typewalk to achieve this.
4 typewalk needs to walk along members, sub and supertypes, this
5 is still missing! irdump then needs to output these edges.
7 ======================================================================
8 ==================== New version: 0.2.5 ==============================
9 ======================================================================
12 Made new Version: 0.2.5
14 15.8.2000 Goetz until 3.9.00
15 Separated irnode.h and many others into two files: one public
17 Added funcionality to extract headers for lib.
20 type_dump also dumps irg->ent
21 if_example: corrected nr of locals
24 irgopt: Correct copy of self pointer in start block.
25 irgopt: Correct treatment of loops: pre function in walker
26 breaks all possible loops in Phi and Block nodes.
27 irr_loop_example: wrong Return statement
28 Now I fixed all but one error: in memory_example.
29 This error showed a conceptual problem, so I wrote
30 a slightly different algorithm. This now runs for
32 dead_node_example: cse is buggy!! merges start and the
36 looking at bug in dead_node_elimination.
38 irnode: all get_..._arr functions changed
39 ircons: new_r_Block: set matured flag.
40 ircons: new_r_SymConst: needs mode_p for linkage_ptr_info.
41 adapted call_str_example to new concepts.
44 Implemented simpler Phi construction algorithm that constructs
45 correct code for three_cfpred_example in ircons. This algorithm
46 generates more Id nodes than the original algorithm and therefore
48 Also made some changes to optimizations of Not.
51 Fixed a bug, caused by the recursive use in type.h and entity.h.
52 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
53 Fixed a problem in if_while testprogram, so it now works.
56 Fixed bug in optimize(). Returned Null instead n if
57 get_optimize() == false.
60 Added some access routines in type.[hc]. Solved a recursive definition
61 of type_class struct, entity struct and type union.
64 Separated entity header into pure functional header and header
68 ... a bunch small changes not documented
71 Changed output of oo_prog_example: now uses the internal list
72 of all irgs. Had to implement some of the access routines for
73 that. Also fixed a bug in irdump. Main routines did not use
74 current_ir_graph properly.
77 Implemented some missing parts in irgopt.
78 Added field irg to entity. Only valid if entity is a method.
79 Had to resove a cyclic de[endency between irgraph.h and entity.h
81 Added fields sub and super to type_class for inheritance.
82 Added field main_irg to irprog.
85 Fixed some bugs in irgopt.c
86 Added some missing 'get_new_node' methods, so the new obstack contains
90 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
91 Solved some copy'n'paste problems. Deleted most debug-code!
92 Added comments to several files.
95 Fixed bug in _r_ constructors for all four Div operations.
96 Fixed bug in turn_into_tuple: didn't remember the block.
97 Tested transform_node.
100 Fixed some testprograms to new semantics of firm.
101 Added testprogram global_var_example.
102 Fixed a bug in irop: Size of Sel node was (entity *) instead of
103 sel_attr. Therefor array_heap_example aborted.
104 Also found a missing break in typewalk, case iro_Alloc.
105 Added a construct that unifies all things constructed by the lib
106 and needed to represent a program: irprog.ch
107 This contains a list of all types as well as a list of all ir graphs.
108 I added calls to initialization and constructors so that this is constructed
109 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
110 Removed strange block struct and constructor from ircons.c.
111 Moved routines for automatic Phi construction from irgmod to ircons.
112 Sorted functions in ircons to reflect the different interfaces.
113 Renamed dataseg to globals to be conformant with documentation.
114 Added conservation of current_ir_graph to irgopt.c.
115 Corrected bug in Makefiles: wrong path to directory ./inlude.
118 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
119 We should change the naming of the Block constructor.
120 Removed acces routines to attr. "closed" of Block.
121 Removed second implementation of new_ir_node from ircons and some comments
122 concerned with the now resolved cyclic dependency.
123 Fixed some bugs in irgopt so that the compilation goes through.
126 Changed the irnode attribut 'visit' into 'visited'.
127 Added some access functions, like get/set_irn_visit etc.
130 Renamed some access funtion in irgraph.*.
131 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
132 Added two attributes to the irgraph struct. The visited flags belongs
133 to a ir graph, not to a whole program.
136 A new optimization has been finished. The dead-node-elimination copies
137 the existing ir graph from one opstack to another. While walking the
138 graph, only the reachable nodes will be copied, so the "dead-nodes"
139 (unreachable in the controlflow) are removed.
142 Solved some problems in the 'copy_node' function. The in array was not
143 as easy as it seamed.
146 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
147 optimization, which is needed in the paramter list of 'irg_walk'.
150 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
151 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
155 Removed some files from the archive, after dependencies and usage are
157 - 'ir/common/strerror.c'
159 - 'ir/ident/xx_ident.h'
160 an empty file, only included by 'ident.c', but never used;
161 - 'ir/tv/label.c' and
163 contained a numeric counter;
164 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
165 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
166 In the old fiasco compiler, the label thing is used in the backend.
167 Search for 'babil' to see the usage.
168 Also removed the two attributes 'lab' and 'used' in the tarval struct.
171 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
172 afterwards some recursive includes, so libfirm works again.
175 Did a lot of changes, which I never commented until now (4.7.00).
177 * Added new result to Start node: Pointer to global data segment.
178 * Extended Semantics of SymConst node to represent information for the
180 * Added arithmeitc nodes (Shrs, Minus ...)
182 Rearranged the directory structure and adjusted the makefiles.
183 The directories contain:
184 ir: everything for the intermediate representation (better: src?)
185 /ir: the ir itself, and standard optimizations.
186 /tv: the target value module
187 /tr: the type and entity representation
188 /adt: abstract data types
189 /common: stuff needed by all other dirs
190 /debug: debugging Unterstuetzung
192 include: external files needen as includes
193 testprograms: examples to test the lib.
195 The makefiles generate files with extension .d that contain the dependencies
199 Added access routine to attribute link of irnode in irnode.ch.
200 Added get_negated_pnc to irnode.ch.
203 Added an iterator in iropt that calls the optimization for all nodes.
204 Copied optimize's code to optimize_in_place and removed deallocations.
205 Made set_irn_n in irnode.c public.
206 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
207 are allowded as cfgpredecessors to represent dead "incoming" control
209 Added call to optimize_in_place in mature_block. Is this useful?
212 Changed tests from comparing enums to comparing pointers. This is more
213 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
214 now get_irn_op == op_And
217 Compared equivalent_value to the one in the original fiasco.
218 Added a test in case iro_And.
219 Changed implementation of iro_Div. Added routines turn_into_tuple
220 and set_irn_op() in irgmod/irnode.
221 Changed implementation of irn_Minus (although commented out).
222 Now all features are ported or documented for this function.
225 Compared computed_value to the one in the original fiasco.
226 Added one further optimization in case iro_Proj.
227 Now all features are ported for this function.
230 Updated comments in ircons.h.
233 Changed possible attribute of SymConst from type_class to type.
234 SymConst size could be the size of a union, array, ...
237 Implemented two dumpers that dump type information and a graph with
241 Implemented a walker that walks over the type informatin that can
242 be reached from an ir node.
243 Added routine get_kind(firm_thing) in common.h, Added file common.c.
244 Added files typewalk.ch and routines to irdump.
245 Added visited flags and initialization to entity and all types.
246 Implemented some of the missing access routines in type.c, entity.c
249 Implemented constructor for ShrA.
251 Added routine that iterates the graph and verifies all nodes.
252 Added Phi_in stack to irgraph. This allows to build several
253 graphs at once with respect to the Phi building algorithm.
256 Implemented dumper for control flow graphs. Changed irr_* and dead_*
257 examples to dump cfg.
260 Implemeted explicit stack for the Phi node generation. See ircons.c.
263 added include string.h in xoprintf.c to get rid of warning:
264 implicit declaration of function `bcopy'
268 Added more cases to equivalent_node.
269 Portet gigo, identify, identify_remember.
270 The set/pset library can not handle several hash tables. At least,
271 compiling several procedures causes a segmentation fault during access
272 to the tables (oo_program_example). Need to work on this.
273 transform_node still needs to be ported.
274 some cases in equivalent_node are still missing.
276 To get it running I had to remove the call to optimize in the
277 new_Block constructor. It can only be called when the Block is
281 Introduced a walker that walks only over blocks. The asserts
282 might be too strict for general graphs, but so far it works.
283 For the walker I introduced an additional flag in the attribute
284 of Blocks, an a global block_visited variable (in irgraph.h).
285 Furhter I modularized dump_ir_node, implemented dump_ir_graph
286 with the existing walker, and added new routines to output a
288 Changed the examples to call the new dump routine.
291 Added auxiliary routines to ir_node:
292 skip_Proj, skip_Nop, is_Bad ...
293 Changed access routines to use get_irn_n instead of explicit array
294 accesses. Changed get_irn_n to call skip_Id, so that the access routines
295 never will return Id nodes.
298 Repaired Phi computation, changed new_r_Phi_in. Added test program to
299 test irregular controlflow.
302 Updated optimizations computed_value and equivalent_node to new
303 Syntax of FIRM nodes and to use access routines.
304 Added routine skip_Proj in irnode.
306 2.2.2000 Boris Boesler
307 Added files irgwalk.c and irgwalk.h. They include a simple function to
308 traverse an ir graph and execute pre and post functions to current node.
311 Added arrays with parameters and result types to the type informtion
312 of a procedure. This is needed for the code generation. Also added
313 the corresponding access routines.
314 Adapted the testprograms.
317 Found a bug / complication: the call order
318 get_value (makes Phi0, put's it into graph_arr)
319 set_value (overwrites Phi0 in graph_arr)
320 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
322 fails. Added test wether graph_arr is already set in phi_merge.
324 Added tons of comments to explain Phi generation -- the stack hack.
328 irdump now outputs the pnc_number for projs of Cmp nodes.
329 Added routine to access the pnc_string.
332 Added result "dataseg" to Start node, with all it's consequences.
333 We need this for the compiler lab to translate global variables.
334 Now they can be selected from the data segment, and the start
335 node supplies the pointer to this segment.
336 These changes are guarded by preprocessor variable UEBPRAKT in
338 Changed files: common.h, irgraph.ch, irnode.h
342 removed array_example, added array-heap_example and
343 array-stack_example and adapted makefile.
345 Entity now also allows type_method as owner. This is to model
346 the stack, Sel nodes that select a variable (array) from the
347 stack need an entity that describes what they select. This
348 entity gets the type of the enclosing procedure as owner.
351 Added future access routines in `irmode.[ch]' as comments,
352 that all attributes in the ir_mode struct can be user-defined.
353 Added access routines in `irgraph.[ch]', for all attributes.
356 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
357 Added `irvrfy.[ch]' to MAKEFILE
358 Added "inline" before access routines in `irnode.h', that no warnings
359 will appear furthermore.
360 Added three more access routines in `irmode.[ch]':
361 struct tarval *get_min_of_mode (ir_mode *mode);
362 struct tarval *get_max_of_mode (ir_mode *mode);
363 struct tarval *get_null_of_mode (ir_mode *mode);
366 Adaption of Alloc, Sel and Entity to new types:
367 Changed array_example to use type instead of type_class.
368 Changed entity to allow all types as owner instead of
369 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
370 Changed alloc to allow allocating variables of any type instead of
372 Adapted all testprograms to avoid warnings.
374 removed dtest ... from Makefile, replaced by test.
377 Added "run" to makefile, added array_example to makefile of testprograms.
378 Corrected testprogram array
379 Changed assert in irvrfy -- Conv.