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