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