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