2 Fixed a bug, caused by the recursive use in type.h and entity.h.
3 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
6 Fixed bug in optimize(). Returned Null instead n if
7 get_optimize() == false.
10 Added some access routines in type.[hc]. Solved a recursive definition
11 of type_class struct, entity struct and type union.
14 Separated entity header into pure functional header and header
18 ... a bunch small changes not documented
21 Changed output of oo_prog_example: now uses the internal list
22 of all irgs. Had to implement some of the access routines for
23 that. Also fixed a bug in irdump. Main routines did not use
24 current_ir_graph properly.
27 Implemented some missing parts in irgopt.
28 Added field irg to entity. Only valid if entity is a method.
29 Had to resove a cyclic de[endency between irgraph.h and entity.h
31 Added fields sub and super to type_class for inheritance.
32 Added field main_irg to irprog.
35 Fixed some bugs in irgopt.c
36 Added some missing 'get_new_node' methods, so the new obstack contains
40 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
41 Solved some copy'n'paste problems. Deleted most debug-code!
42 Added comments to several files.
45 Fixed bug in _r_ constructors for all four Div operations.
46 Fixed bug in turn_into_tuple: didn't remember the block.
47 Tested transform_node.
50 Fixed some testprograms to new semantics of firm.
51 Added testprogram global_var_example.
52 Fixed a bug in irop: Size of Sel node was (entity *) instead of
53 sel_attr. Therefor array_heap_example aborted.
54 Also found a missing break in typewalk, case iro_Alloc.
55 Added a construct that unifies all things constructed by the lib
56 and needed to represent a program: irprog.ch
57 This contains a list of all types as well as a list of all ir graphs.
58 I added calls to initialization and constructors so that this is constructed
59 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
60 Removed strange block struct and constructor from ircons.c.
61 Moved routines for automatic Phi construction from irgmod to ircons.
62 Sorted functions in ircons to reflect the different interfaces.
63 Renamed dataseg to globals to be conformant with documentation.
64 Added conservation of current_ir_graph to irgopt.c.
65 Corrected bug in Makefiles: wrong path to directory ./inlude.
68 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
69 We should change the naming of the Block constructor.
70 Removed acces routines to attr. "closed" of Block.
71 Removed second implementation of new_ir_node from ircons and some comments
72 concerned with the now resolved cyclic dependency.
73 Fixed some bugs in irgopt so that the compilation goes through.
76 Changed the irnode attribut 'visit' into 'visited'.
77 Added some access functions, like get/set_irn_visit etc.
80 Renamed some access funtion in irgraph.*.
81 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
82 Added two attributes to the irgraph struct. The visited flags belongs
83 to a ir graph, not to a whole program.
86 A new optimization has been finished. The dead-node-elimination copies
87 the existing ir graph from one opstack to another. While walking the
88 graph, only the reachable nodes will be copied, so the "dead-nodes"
89 (unreachable in the controlflow) are removed.
92 Solved some problems in the 'copy_node' function. The in array was not
96 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
97 optimization, which is needed in the paramter list of 'irg_walk'.
100 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
101 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
105 Removed some files from the archive, after dependencies and usage are
107 - 'ir/common/strerror.c'
109 - 'ir/ident/xx_ident.h'
110 an empty file, only included by 'ident.c', but never used;
111 - 'ir/tv/label.c' and
113 contained a numeric counter;
114 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
115 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
116 In the old fiasco compiler, the label thing is used in the backend.
117 Search for 'babil' to see the usage.
118 Also removed the two attributes 'lab' and 'used' in the tarval struct.
121 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
122 afterwards some recursive includes, so libfirm works again.
125 Did a lot of changes, which I never commented until now (4.7.00).
127 * Added new result to Start node: Pointer to global data segment.
128 * Extended Semantics of SymConst node to represent information for the
130 * Added arithmeitc nodes (Shrs, Minus ...)
132 Rearranged the directory structure and adjusted the makefiles.
133 The directories contain:
134 ir: everything for the intermediate representation (better: src?)
135 /ir: the ir itself, and standard optimizations.
136 /tv: the target value module
137 /tr: the type and entity representation
138 /adt: abstract data types
139 /common: stuff needed by all other dirs
140 /debug: debugging Unterstuetzung
142 include: external files needen as includes
143 testprograms: examples to test the lib.
145 The makefiles generate files with extension .d that contain the dependencies
149 Added access routine to attribute link of irnode in irnode.ch.
150 Added get_negated_pnc to irnode.ch.
153 Added an iterator in iropt that calls the optimization for all nodes.
154 Copied optimize's code to optimize_in_place and removed deallocations.
155 Made set_irn_n in irnode.c public.
156 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
157 are allowded as cfgpredecessors to represent dead "incoming" control
159 Added call to optimize_in_place in mature_block. Is this useful?
162 Changed tests from comparing enums to comparing pointers. This is more
163 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
164 now get_irn_op == op_And
167 Compared equivalent_value to the one in the original fiasco.
168 Added a test in case iro_And.
169 Changed implementation of iro_Div. Added routines turn_into_tuple
170 and set_irn_op() in irgmod/irnode.
171 Changed implementation of irn_Minus (although commented out).
172 Now all features are ported or documented for this function.
175 Compared computed_value to the one in the original fiasco.
176 Added one further optimization in case iro_Proj.
177 Now all features are ported for this function.
180 Updated comments in ircons.h.
183 Changed possible attribute of SymConst from type_class to type.
184 SymConst size could be the size of a union, array, ...
187 Implemented two dumpers that dump type information and a graph with
191 Implemented a walker that walks over the type informatin that can
192 be reached from an ir node.
193 Added routine get_kind(firm_thing) in common.h, Added file common.c.
194 Added files typewalk.ch and routines to irdump.
195 Added visited flags and initialization to entity and all types.
196 Implemented some of the missing access routines in type.c, entity.c
199 Implemented constructor for ShrA.
201 Added routine that iterates the graph and verifies all nodes.
202 Added Phi_in stack to irgraph. This allows to build several
203 graphs at once with respect to the Phi building algorithm.
206 Implemented dumper for control flow graphs. Changed irr_* and dead_*
207 examples to dump cfg.
210 Implemeted explicit stack for the Phi node generation. See ircons.c.
213 added include string.h in xoprintf.c to get rid of warning:
214 implicit declaration of function `bcopy'
218 Added more cases to equivalent_node.
219 Portet gigo, identify, identify_remember.
220 The set/pset library can not handle several hash tables. At least,
221 compiling several procedures causes a segmentation fault during access
222 to the tables (oo_program_example). Need to work on this.
223 transform_node still needs to be ported.
224 some cases in equivalent_node are still missing.
226 To get it running I had to remove the call to optimize in the
227 new_Block constructor. It can only be called when the Block is
231 Introduced a walker that walks only over blocks. The asserts
232 might be too strict for general graphs, but so far it works.
233 For the walker I introduced an additional flag in the attribute
234 of Blocks, an a global block_visited variable (in irgraph.h).
235 Furhter I modularized dump_ir_node, implemented dump_ir_graph
236 with the existing walker, and added new routines to output a
238 Changed the examples to call the new dump routine.
241 Added auxiliary routines to ir_node:
242 skip_Proj, skip_Nop, is_Bad ...
243 Changed access routines to use get_irn_n instead of explicit array
244 accesses. Changed get_irn_n to call skip_Id, so that the access routines
245 never will return Id nodes.
248 Repaired Phi computation, changed new_r_Phi_in. Added test program to
249 test irregular controlflow.
252 Updated optimizations computed_value and equivalent_node to new
253 Syntax of FIRM nodes and to use access routines.
254 Added routine skip_Proj in irnode.
256 2.2.2000 Boris Boesler
257 Added files irgwalk.c and irgwalk.h. They include a simple function to
258 traverse an ir graph and execute pre and post functions to current node.
261 Added arrays with parameters and result types to the type informtion
262 of a procedure. This is needed for the code generation. Also added
263 the corresponding access routines.
264 Adapted the testprograms.
267 Found a bug / complication: the call order
268 get_value (makes Phi0, put's it into graph_arr)
269 set_value (overwrites Phi0 in graph_arr)
270 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
272 fails. Added test wether graph_arr is already set in phi_merge.
274 Added tons of comments to explain Phi generation -- the stack hack.
278 irdump now outputs the pnc_number for projs of Cmp nodes.
279 Added routine to access the pnc_string.
282 Added result "dataseg" to Start node, with all it's consequences.
283 We need this for the compiler lab to translate global variables.
284 Now they can be selected from the data segment, and the start
285 node supplies the pointer to this segment.
286 These changes are guarded by preprocessor variable UEBPRAKT in
288 Changed files: common.h, irgraph.ch, irnode.h
292 removed array_example, added array-heap_example and
293 array-stack_example and adapted makefile.
295 Entity now also allows type_method as owner. This is to model
296 the stack, Sel nodes that select a variable (array) from the
297 stack need an entity that describes what they select. This
298 entity gets the type of the enclosing procedure as owner.
301 Added future access routines in `irmode.[ch]' as comments,
302 that all attributes in the ir_mode struct can be user-defined.
303 Added access routines in `irgraph.[ch]', for all attributes.
306 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
307 Added `irvrfy.[ch]' to MAKEFILE
308 Added "inline" before access routines in `irnode.h', that no warnings
309 will appear furthermore.
310 Added three more access routines in `irmode.[ch]':
311 struct tarval *get_min_of_mode (ir_mode *mode);
312 struct tarval *get_max_of_mode (ir_mode *mode);
313 struct tarval *get_null_of_mode (ir_mode *mode);
316 Adaption of Alloc, Sel and Entity to new types:
317 Changed array_example to use type instead of type_class.
318 Changed entity to allow all types as owner instead of
319 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
320 Changed alloc to allow allocating variables of any type instead of
322 Adapted all testprograms to avoid warnings.
324 removed dtest ... from Makefile, replaced by test.
327 Added "run" to makefile, added array_example to makefile of testprograms.
328 Corrected testprogram array
329 Changed assert in irvrfy -- Conv.