2 Changed Makefile and Configure scripts to produce a proper tar file
3 for distribution (libfirm-VERSION.tar)
4 Use 'make distrib' instead of 'make dist'
7 Changed implementation of tr module. With this I also changed the
8 interface!! Type casts are unnecessary now, but some access funcions
12 ======================================================================
13 ==================== New version: 0.3.0 ==============================
14 ======================================================================
20 added creation of a tar archive for distribution
21 replaced ir/.dummy.in by ir/Makefile.in
24 Changed equivalent_node() and copy_preds() so that the end block
25 is not optimized away if it has no real predecessors. This fixes
26 the problem with the while examples.
27 Added while_example in Makefile
30 Fooled around with if_while_example: Why does cse make it abort?
31 Cse removes path to End node, Bad remains as pred of Endblock.
32 The block walker can't deal with this.
33 Made new example with only loop: while_example.
36 added auto documentation in Makefiles, configure, etc
37 example in ir/ir/ircons.c - much must be added
38 creation of documentation by 'make autodoc' in architecture directory
41 added lots of #ifdef HAVE_CONFIG_G #include "config.h" #endif
42 replaced some C++ style comments (//) by standard C comments /**/
45 - configure (with some features, see --help for more)
46 - make [depend | tags | install]
47 - make [run] in testprograms
49 - compilation on SunOS
51 - automatic call of autoconf if required
54 More colours in vcg output.
55 Fixed bug: cse found that start block equals it's
56 first successor as both have the Start block as predecessor.
57 This only happens for local_optimize_graph() as during construction
58 the Start block is not entered in the value table.
59 Changed implementation: irgmod: local_optimize_graph() did not
60 empty the hash table with the irnodes for cse. This might cause that
61 cse turns up with a node that seems better but was removed from the
62 graph in an earlier optimization phase.
63 Removed old #if 1 from ircons.c
64 Commented in call of optimization for Block nodes in irgopt.c, copy_preds():
65 Optimizes blocks that had bad predecessors.
70 Moved struct from irprog.h to irprog_t.h, same for irmode.h.
71 Added a module deb_info in debug.h. Adapted makefile. Added example
73 Removed use of debug.h in ident.c. Now debug.c|h can be moved to
75 Removed inabled code for dead node elimination in irgopt.h.
76 Renamed some missnamed access routines (_of_). Macros with old names
77 are in old_fctnames.h.
78 Edited makefiles to remove #* and *.flc.
81 Some minor bugfixes...
84 After compacting of in arrays in dead_node_elimination
85 merge blocks and remove Phis with one pred.
88 Freeing of obstack in dead_node_elimination commented in.
89 Compacting in arrays of Block and Phi nodes in dead_node_
94 added set/get_SymConst_type_or_id in irnode.h
95 added field offset to entity in entity.h
96 added field size to type_class in type.h
97 include "gmp.h" in tv.h with "" instead <>
100 Some changes to make the lib better portable
103 Initialize max_node_nr in irprog.
106 Now outputs warning if faulty type in irdump, dump_type_info.
107 Explicit string termination in irdump, vcg_open.
110 Added external to declaration of irp.
111 Added typedef before pns and pnc_number.
112 irmode.c, init_mode(): We need to memset all new modes as
113 tarval_vrfy reads fields before they are initialized: It
114 compares to min/max when tarvals for min/max are allocated!
117 Changed parameter of Methods in type.h and type.c from
118 class to clss. Class is a C++ keyword.
121 Changed some testprograms to accomodate Uebprakt.
124 Finished implementation of dump_all_types.
125 Added new testprogram inheritance_example.
128 Added seven access routines in type.[hc]:
129 get_class_n_member, get_class_n_subtype, get_class_n_supertype,
130 get_strct_n_member, add_strct_member, get_strct_member and set_strct_member.
131 Fixed the type_walk_2 for those seven routines.
134 Added irdump routine dump_all_types.
135 Did changes to irprog, type, typewalk to achieve this.
136 typewalk needs to walk along members, sub and supertypes, this
137 is still missing! irdump then needs to output these edges.
139 ======================================================================
140 ==================== New version: 0.2.5 ==============================
141 ======================================================================
144 Made new Version: 0.2.5
146 15.8.2000 Goetz until 3.9.00
147 Separated irnode.h and many others into two files: one public
149 Added funcionality to extract headers for lib.
152 type_dump also dumps irg->ent
153 if_example: corrected nr of locals
156 irgopt: Correct copy of self pointer in start block.
157 irgopt: Correct treatment of loops: pre function in walker
158 breaks all possible loops in Phi and Block nodes.
159 irr_loop_example: wrong Return statement
160 Now I fixed all but one error: in memory_example.
161 This error showed a conceptual problem, so I wrote
162 a slightly different algorithm. This now runs for
164 dead_node_example: cse is buggy!! merges start and the
168 looking at bug in dead_node_elimination.
170 irnode: all get_..._arr functions changed
171 ircons: new_r_Block: set matured flag.
172 ircons: new_r_SymConst: needs mode_p for linkage_ptr_info.
173 adapted call_str_example to new concepts.
176 Implemented simpler Phi construction algorithm that constructs
177 correct code for three_cfpred_example in ircons. This algorithm
178 generates more Id nodes than the original algorithm and therefore
180 Also made some changes to optimizations of Not.
183 Fixed a bug, caused by the recursive use in type.h and entity.h.
184 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
185 Fixed a problem in if_while testprogram, so it now works.
188 Fixed bug in optimize(). Returned Null instead n if
189 get_optimize() == false.
192 Added some access routines in type.[hc]. Solved a recursive definition
193 of type_class struct, entity struct and type union.
196 Separated entity header into pure functional header and header
200 ... a bunch small changes not documented
203 Changed output of oo_prog_example: now uses the internal list
204 of all irgs. Had to implement some of the access routines for
205 that. Also fixed a bug in irdump. Main routines did not use
206 current_ir_graph properly.
209 Implemented some missing parts in irgopt.
210 Added field irg to entity. Only valid if entity is a method.
211 Had to resove a cyclic de[endency between irgraph.h and entity.h
213 Added fields sub and super to type_class for inheritance.
214 Added field main_irg to irprog.
217 Fixed some bugs in irgopt.c
218 Added some missing 'get_new_node' methods, so the new obstack contains
222 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
223 Solved some copy'n'paste problems. Deleted most debug-code!
224 Added comments to several files.
227 Fixed bug in _r_ constructors for all four Div operations.
228 Fixed bug in turn_into_tuple: didn't remember the block.
229 Tested transform_node.
232 Fixed some testprograms to new semantics of firm.
233 Added testprogram global_var_example.
234 Fixed a bug in irop: Size of Sel node was (entity *) instead of
235 sel_attr. Therefor array_heap_example aborted.
236 Also found a missing break in typewalk, case iro_Alloc.
237 Added a construct that unifies all things constructed by the lib
238 and needed to represent a program: irprog.ch
239 This contains a list of all types as well as a list of all ir graphs.
240 I added calls to initialization and constructors so that this is constructed
241 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
242 Removed strange block struct and constructor from ircons.c.
243 Moved routines for automatic Phi construction from irgmod to ircons.
244 Sorted functions in ircons to reflect the different interfaces.
245 Renamed dataseg to globals to be conformant with documentation.
246 Added conservation of current_ir_graph to irgopt.c.
247 Corrected bug in Makefiles: wrong path to directory ./inlude.
250 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
251 We should change the naming of the Block constructor.
252 Removed acces routines to attr. "closed" of Block.
253 Removed second implementation of new_ir_node from ircons and some comments
254 concerned with the now resolved cyclic dependency.
255 Fixed some bugs in irgopt so that the compilation goes through.
258 Changed the irnode attribut 'visit' into 'visited'.
259 Added some access functions, like get/set_irn_visit etc.
262 Renamed some access funtion in irgraph.*.
263 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
264 Added two attributes to the irgraph struct. The visited flags belongs
265 to a ir graph, not to a whole program.
268 A new optimization has been finished. The dead-node-elimination copies
269 the existing ir graph from one opstack to another. While walking the
270 graph, only the reachable nodes will be copied, so the "dead-nodes"
271 (unreachable in the controlflow) are removed.
274 Solved some problems in the 'copy_node' function. The in array was not
275 as easy as it seamed.
278 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
279 optimization, which is needed in the paramter list of 'irg_walk'.
282 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
283 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
287 Removed some files from the archive, after dependencies and usage are
289 - 'ir/common/strerror.c'
291 - 'ir/ident/xx_ident.h'
292 an empty file, only included by 'ident.c', but never used;
293 - 'ir/tv/label.c' and
295 contained a numeric counter;
296 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
297 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
298 In the old fiasco compiler, the label thing is used in the backend.
299 Search for 'babil' to see the usage.
300 Also removed the two attributes 'lab' and 'used' in the tarval struct.
303 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
304 afterwards some recursive includes, so libfirm works again.
307 Did a lot of changes, which I never commented until now (4.7.00).
309 * Added new result to Start node: Pointer to global data segment.
310 * Extended Semantics of SymConst node to represent information for the
312 * Added arithmeitc nodes (Shrs, Minus ...)
314 Rearranged the directory structure and adjusted the makefiles.
315 The directories contain:
316 ir: everything for the intermediate representation (better: src?)
317 /ir: the ir itself, and standard optimizations.
318 /tv: the target value module
319 /tr: the type and entity representation
320 /adt: abstract data types
321 /common: stuff needed by all other dirs
322 /debug: debugging Unterstuetzung
324 include: external files needen as includes
325 testprograms: examples to test the lib.
327 The makefiles generate files with extension .d that contain the dependencies
331 Added access routine to attribute link of irnode in irnode.ch.
332 Added get_negated_pnc to irnode.ch.
335 Added an iterator in iropt that calls the optimization for all nodes.
336 Copied optimize's code to optimize_in_place and removed deallocations.
337 Made set_irn_n in irnode.c public.
338 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
339 are allowded as cfgpredecessors to represent dead "incoming" control
341 Added call to optimize_in_place in mature_block. Is this useful?
344 Changed tests from comparing enums to comparing pointers. This is more
345 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
346 now get_irn_op == op_And
349 Compared equivalent_value to the one in the original fiasco.
350 Added a test in case iro_And.
351 Changed implementation of iro_Div. Added routines turn_into_tuple
352 and set_irn_op() in irgmod/irnode.
353 Changed implementation of irn_Minus (although commented out).
354 Now all features are ported or documented for this function.
357 Compared computed_value to the one in the original fiasco.
358 Added one further optimization in case iro_Proj.
359 Now all features are ported for this function.
362 Updated comments in ircons.h.
365 Changed possible attribute of SymConst from type_class to type.
366 SymConst size could be the size of a union, array, ...
369 Implemented two dumpers that dump type information and a graph with
373 Implemented a walker that walks over the type informatin that can
374 be reached from an ir node.
375 Added routine get_kind(firm_thing) in common.h, Added file common.c.
376 Added files typewalk.ch and routines to irdump.
377 Added visited flags and initialization to entity and all types.
378 Implemented some of the missing access routines in type.c, entity.c
381 Implemented constructor for ShrA.
383 Added routine that iterates the graph and verifies all nodes.
384 Added Phi_in stack to irgraph. This allows to build several
385 graphs at once with respect to the Phi building algorithm.
388 Implemented dumper for control flow graphs. Changed irr_* and dead_*
389 examples to dump cfg.
392 Implemeted explicit stack for the Phi node generation. See ircons.c.
395 added include string.h in xoprintf.c to get rid of warning:
396 implicit declaration of function `bcopy'
400 Added more cases to equivalent_node.
401 Portet gigo, identify, identify_remember.
402 The set/pset library can not handle several hash tables. At least,
403 compiling several procedures causes a segmentation fault during access
404 to the tables (oo_program_example). Need to work on this.
405 transform_node still needs to be ported.
406 some cases in equivalent_node are still missing.
408 To get it running I had to remove the call to optimize in the
409 new_Block constructor. It can only be called when the Block is
413 Introduced a walker that walks only over blocks. The asserts
414 might be too strict for general graphs, but so far it works.
415 For the walker I introduced an additional flag in the attribute
416 of Blocks, an a global block_visited variable (in irgraph.h).
417 Furhter I modularized dump_ir_node, implemented dump_ir_graph
418 with the existing walker, and added new routines to output a
420 Changed the examples to call the new dump routine.
423 Added auxiliary routines to ir_node:
424 skip_Proj, skip_Nop, is_Bad ...
425 Changed access routines to use get_irn_n instead of explicit array
426 accesses. Changed get_irn_n to call skip_Id, so that the access routines
427 never will return Id nodes.
430 Repaired Phi computation, changed new_r_Phi_in. Added test program to
431 test irregular controlflow.
434 Updated optimizations computed_value and equivalent_node to new
435 Syntax of FIRM nodes and to use access routines.
436 Added routine skip_Proj in irnode.
438 2.2.2000 Boris Boesler
439 Added files irgwalk.c and irgwalk.h. They include a simple function to
440 traverse an ir graph and execute pre and post functions to current node.
443 Added arrays with parameters and result types to the type informtion
444 of a procedure. This is needed for the code generation. Also added
445 the corresponding access routines.
446 Adapted the testprograms.
449 Found a bug / complication: the call order
450 get_value (makes Phi0, put's it into graph_arr)
451 set_value (overwrites Phi0 in graph_arr)
452 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
454 fails. Added test wether graph_arr is already set in phi_merge.
456 Added tons of comments to explain Phi generation -- the stack hack.
460 irdump now outputs the pnc_number for projs of Cmp nodes.
461 Added routine to access the pnc_string.
464 Added result "dataseg" to Start node, with all it's consequences.
465 We need this for the compiler lab to translate global variables.
466 Now they can be selected from the data segment, and the start
467 node supplies the pointer to this segment.
468 These changes are guarded by preprocessor variable UEBPRAKT in
470 Changed files: common.h, irgraph.ch, irnode.h
474 removed array_example, added array-heap_example and
475 array-stack_example and adapted makefile.
477 Entity now also allows type_method as owner. This is to model
478 the stack, Sel nodes that select a variable (array) from the
479 stack need an entity that describes what they select. This
480 entity gets the type of the enclosing procedure as owner.
483 Added future access routines in `irmode.[ch]' as comments,
484 that all attributes in the ir_mode struct can be user-defined.
485 Added access routines in `irgraph.[ch]', for all attributes.
488 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
489 Added `irvrfy.[ch]' to MAKEFILE
490 Added "inline" before access routines in `irnode.h', that no warnings
491 will appear furthermore.
492 Added three more access routines in `irmode.[ch]':
493 struct tarval *get_min_of_mode (ir_mode *mode);
494 struct tarval *get_max_of_mode (ir_mode *mode);
495 struct tarval *get_null_of_mode (ir_mode *mode);
498 Adaption of Alloc, Sel and Entity to new types:
499 Changed array_example to use type instead of type_class.
500 Changed entity to allow all types as owner instead of
501 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
502 Changed alloc to allow allocating variables of any type instead of
504 Adapted all testprograms to avoid warnings.
506 removed dtest ... from Makefile, replaced by test.
509 Added "run" to makefile, added array_example to makefile of testprograms.
510 Corrected testprogram array
511 Changed assert in irvrfy -- Conv.