2 Changed Makefile/MakeTargets to install headers and libraries
6 Changed Makefile and Configure scripts to produce a proper tar file
7 for distribution (libfirm-VERSION.tar)
8 Use 'make distrib' instead of 'make dist'
11 Changed implementation of tr module. With this I also changed the
12 interface!! Type casts are unnecessary now, but some access funcions
16 ======================================================================
17 ==================== New version: 0.3.0 ==============================
18 ======================================================================
24 added creation of a tar archive for distribution
25 replaced ir/.dummy.in by ir/Makefile.in
28 Changed equivalent_node() and copy_preds() so that the end block
29 is not optimized away if it has no real predecessors. This fixes
30 the problem with the while examples.
31 Added while_example in Makefile
34 Fooled around with if_while_example: Why does cse make it abort?
35 Cse removes path to End node, Bad remains as pred of Endblock.
36 The block walker can't deal with this.
37 Made new example with only loop: while_example.
40 added auto documentation in Makefiles, configure, etc
41 example in ir/ir/ircons.c - much must be added
42 creation of documentation by 'make autodoc' in architecture directory
45 added lots of #ifdef HAVE_CONFIG_G #include "config.h" #endif
46 replaced some C++ style comments (//) by standard C comments /**/
49 - configure (with some features, see --help for more)
50 - make [depend | tags | install]
51 - make [run] in testprograms
53 - compilation on SunOS
55 - automatic call of autoconf if required
58 More colours in vcg output.
59 Fixed bug: cse found that start block equals it's
60 first successor as both have the Start block as predecessor.
61 This only happens for local_optimize_graph() as during construction
62 the Start block is not entered in the value table.
63 Changed implementation: irgmod: local_optimize_graph() did not
64 empty the hash table with the irnodes for cse. This might cause that
65 cse turns up with a node that seems better but was removed from the
66 graph in an earlier optimization phase.
67 Removed old #if 1 from ircons.c
68 Commented in call of optimization for Block nodes in irgopt.c, copy_preds():
69 Optimizes blocks that had bad predecessors.
74 Moved struct from irprog.h to irprog_t.h, same for irmode.h.
75 Added a module deb_info in debug.h. Adapted makefile. Added example
77 Removed use of debug.h in ident.c. Now debug.c|h can be moved to
79 Removed inabled code for dead node elimination in irgopt.h.
80 Renamed some missnamed access routines (_of_). Macros with old names
81 are in old_fctnames.h.
82 Edited makefiles to remove #* and *.flc.
85 Some minor bugfixes...
88 After compacting of in arrays in dead_node_elimination
89 merge blocks and remove Phis with one pred.
92 Freeing of obstack in dead_node_elimination commented in.
93 Compacting in arrays of Block and Phi nodes in dead_node_
98 added set/get_SymConst_type_or_id in irnode.h
99 added field offset to entity in entity.h
100 added field size to type_class in type.h
101 include "gmp.h" in tv.h with "" instead <>
104 Some changes to make the lib better portable
107 Initialize max_node_nr in irprog.
110 Now outputs warning if faulty type in irdump, dump_type_info.
111 Explicit string termination in irdump, vcg_open.
114 Added external to declaration of irp.
115 Added typedef before pns and pnc_number.
116 irmode.c, init_mode(): We need to memset all new modes as
117 tarval_vrfy reads fields before they are initialized: It
118 compares to min/max when tarvals for min/max are allocated!
121 Changed parameter of Methods in type.h and type.c from
122 class to clss. Class is a C++ keyword.
125 Changed some testprograms to accomodate Uebprakt.
128 Finished implementation of dump_all_types.
129 Added new testprogram inheritance_example.
132 Added seven access routines in type.[hc]:
133 get_class_n_member, get_class_n_subtype, get_class_n_supertype,
134 get_strct_n_member, add_strct_member, get_strct_member and set_strct_member.
135 Fixed the type_walk_2 for those seven routines.
138 Added irdump routine dump_all_types.
139 Did changes to irprog, type, typewalk to achieve this.
140 typewalk needs to walk along members, sub and supertypes, this
141 is still missing! irdump then needs to output these edges.
143 ======================================================================
144 ==================== New version: 0.2.5 ==============================
145 ======================================================================
148 Made new Version: 0.2.5
150 15.8.2000 Goetz until 3.9.00
151 Separated irnode.h and many others into two files: one public
153 Added funcionality to extract headers for lib.
156 type_dump also dumps irg->ent
157 if_example: corrected nr of locals
160 irgopt: Correct copy of self pointer in start block.
161 irgopt: Correct treatment of loops: pre function in walker
162 breaks all possible loops in Phi and Block nodes.
163 irr_loop_example: wrong Return statement
164 Now I fixed all but one error: in memory_example.
165 This error showed a conceptual problem, so I wrote
166 a slightly different algorithm. This now runs for
168 dead_node_example: cse is buggy!! merges start and the
172 looking at bug in dead_node_elimination.
174 irnode: all get_..._arr functions changed
175 ircons: new_r_Block: set matured flag.
176 ircons: new_r_SymConst: needs mode_p for linkage_ptr_info.
177 adapted call_str_example to new concepts.
180 Implemented simpler Phi construction algorithm that constructs
181 correct code for three_cfpred_example in ircons. This algorithm
182 generates more Id nodes than the original algorithm and therefore
184 Also made some changes to optimizations of Not.
187 Fixed a bug, caused by the recursive use in type.h and entity.h.
188 Deleted the _TYPE_TYPEDEF_ and inserted the _ENTITY_TYPEDEF_ instead.
189 Fixed a problem in if_while testprogram, so it now works.
192 Fixed bug in optimize(). Returned Null instead n if
193 get_optimize() == false.
196 Added some access routines in type.[hc]. Solved a recursive definition
197 of type_class struct, entity struct and type union.
200 Separated entity header into pure functional header and header
204 ... a bunch small changes not documented
207 Changed output of oo_prog_example: now uses the internal list
208 of all irgs. Had to implement some of the access routines for
209 that. Also fixed a bug in irdump. Main routines did not use
210 current_ir_graph properly.
213 Implemented some missing parts in irgopt.
214 Added field irg to entity. Only valid if entity is a method.
215 Had to resove a cyclic de[endency between irgraph.h and entity.h
217 Added fields sub and super to type_class for inheritance.
218 Added field main_irg to irprog.
221 Fixed some bugs in irgopt.c
222 Added some missing 'get_new_node' methods, so the new obstack contains
226 Renamed ir_vrfy to irn_vrfy and vrfy_graphg to irg_vrfy.
227 Solved some copy'n'paste problems. Deleted most debug-code!
228 Added comments to several files.
231 Fixed bug in _r_ constructors for all four Div operations.
232 Fixed bug in turn_into_tuple: didn't remember the block.
233 Tested transform_node.
236 Fixed some testprograms to new semantics of firm.
237 Added testprogram global_var_example.
238 Fixed a bug in irop: Size of Sel node was (entity *) instead of
239 sel_attr. Therefor array_heap_example aborted.
240 Also found a missing break in typewalk, case iro_Alloc.
241 Added a construct that unifies all things constructed by the lib
242 and needed to represent a program: irprog.ch
243 This contains a list of all types as well as a list of all ir graphs.
244 I added calls to initialization and constructors so that this is constructed
245 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
246 Removed strange block struct and constructor from ircons.c.
247 Moved routines for automatic Phi construction from irgmod to ircons.
248 Sorted functions in ircons to reflect the different interfaces.
249 Renamed dataseg to globals to be conformant with documentation.
250 Added conservation of current_ir_graph to irgopt.c.
251 Corrected bug in Makefiles: wrong path to directory ./inlude.
254 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
255 We should change the naming of the Block constructor.
256 Removed acces routines to attr. "closed" of Block.
257 Removed second implementation of new_ir_node from ircons and some comments
258 concerned with the now resolved cyclic dependency.
259 Fixed some bugs in irgopt so that the compilation goes through.
262 Changed the irnode attribut 'visit' into 'visited'.
263 Added some access functions, like get/set_irn_visit etc.
266 Renamed some access funtion in irgraph.*.
267 Deleted two global flags (ir_visited and block_visited) from irgraph.*.
268 Added two attributes to the irgraph struct. The visited flags belongs
269 to a ir graph, not to a whole program.
272 A new optimization has been finished. The dead-node-elimination copies
273 the existing ir graph from one opstack to another. While walking the
274 graph, only the reachable nodes will be copied, so the "dead-nodes"
275 (unreachable in the controlflow) are removed.
278 Solved some problems in the 'copy_node' function. The in array was not
279 as easy as it seamed.
282 Wrote a new 'post' funtion, for future use in 'dead-node-elimination'
283 optimization, which is needed in the paramter list of 'irg_walk'.
286 Moved 'optimze_in_place_wrapper' from iropt.* tothe new files irgopt.*.
287 New functions 'local_optimze_graph' and 'dead_node_elimination' are added,
291 Removed some files from the archive, after dependencies and usage are
293 - 'ir/common/strerror.c'
295 - 'ir/ident/xx_ident.h'
296 an empty file, only included by 'ident.c', but never used;
297 - 'ir/tv/label.c' and
299 contained a numeric counter;
300 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
301 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
302 In the old fiasco compiler, the label thing is used in the backend.
303 Search for 'babil' to see the usage.
304 Also removed the two attributes 'lab' and 'used' in the tarval struct.
307 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
308 afterwards some recursive includes, so libfirm works again.
311 Did a lot of changes, which I never commented until now (4.7.00).
313 * Added new result to Start node: Pointer to global data segment.
314 * Extended Semantics of SymConst node to represent information for the
316 * Added arithmeitc nodes (Shrs, Minus ...)
318 Rearranged the directory structure and adjusted the makefiles.
319 The directories contain:
320 ir: everything for the intermediate representation (better: src?)
321 /ir: the ir itself, and standard optimizations.
322 /tv: the target value module
323 /tr: the type and entity representation
324 /adt: abstract data types
325 /common: stuff needed by all other dirs
326 /debug: debugging Unterstuetzung
328 include: external files needen as includes
329 testprograms: examples to test the lib.
331 The makefiles generate files with extension .d that contain the dependencies
335 Added access routine to attribute link of irnode in irnode.ch.
336 Added get_negated_pnc to irnode.ch.
339 Added an iterator in iropt that calls the optimization for all nodes.
340 Copied optimize's code to optimize_in_place and removed deallocations.
341 Made set_irn_n in irnode.c public.
342 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
343 are allowded as cfgpredecessors to represent dead "incoming" control
345 Added call to optimize_in_place in mature_block. Is this useful?
348 Changed tests from comparing enums to comparing pointers. This is more
349 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
350 now get_irn_op == op_And
353 Compared equivalent_value to the one in the original fiasco.
354 Added a test in case iro_And.
355 Changed implementation of iro_Div. Added routines turn_into_tuple
356 and set_irn_op() in irgmod/irnode.
357 Changed implementation of irn_Minus (although commented out).
358 Now all features are ported or documented for this function.
361 Compared computed_value to the one in the original fiasco.
362 Added one further optimization in case iro_Proj.
363 Now all features are ported for this function.
366 Updated comments in ircons.h.
369 Changed possible attribute of SymConst from type_class to type.
370 SymConst size could be the size of a union, array, ...
373 Implemented two dumpers that dump type information and a graph with
377 Implemented a walker that walks over the type informatin that can
378 be reached from an ir node.
379 Added routine get_kind(firm_thing) in common.h, Added file common.c.
380 Added files typewalk.ch and routines to irdump.
381 Added visited flags and initialization to entity and all types.
382 Implemented some of the missing access routines in type.c, entity.c
385 Implemented constructor for ShrA.
387 Added routine that iterates the graph and verifies all nodes.
388 Added Phi_in stack to irgraph. This allows to build several
389 graphs at once with respect to the Phi building algorithm.
392 Implemented dumper for control flow graphs. Changed irr_* and dead_*
393 examples to dump cfg.
396 Implemeted explicit stack for the Phi node generation. See ircons.c.
399 added include string.h in xoprintf.c to get rid of warning:
400 implicit declaration of function `bcopy'
404 Added more cases to equivalent_node.
405 Portet gigo, identify, identify_remember.
406 The set/pset library can not handle several hash tables. At least,
407 compiling several procedures causes a segmentation fault during access
408 to the tables (oo_program_example). Need to work on this.
409 transform_node still needs to be ported.
410 some cases in equivalent_node are still missing.
412 To get it running I had to remove the call to optimize in the
413 new_Block constructor. It can only be called when the Block is
417 Introduced a walker that walks only over blocks. The asserts
418 might be too strict for general graphs, but so far it works.
419 For the walker I introduced an additional flag in the attribute
420 of Blocks, an a global block_visited variable (in irgraph.h).
421 Furhter I modularized dump_ir_node, implemented dump_ir_graph
422 with the existing walker, and added new routines to output a
424 Changed the examples to call the new dump routine.
427 Added auxiliary routines to ir_node:
428 skip_Proj, skip_Nop, is_Bad ...
429 Changed access routines to use get_irn_n instead of explicit array
430 accesses. Changed get_irn_n to call skip_Id, so that the access routines
431 never will return Id nodes.
434 Repaired Phi computation, changed new_r_Phi_in. Added test program to
435 test irregular controlflow.
438 Updated optimizations computed_value and equivalent_node to new
439 Syntax of FIRM nodes and to use access routines.
440 Added routine skip_Proj in irnode.
442 2.2.2000 Boris Boesler
443 Added files irgwalk.c and irgwalk.h. They include a simple function to
444 traverse an ir graph and execute pre and post functions to current node.
447 Added arrays with parameters and result types to the type informtion
448 of a procedure. This is needed for the code generation. Also added
449 the corresponding access routines.
450 Adapted the testprograms.
453 Found a bug / complication: the call order
454 get_value (makes Phi0, put's it into graph_arr)
455 set_value (overwrites Phi0 in graph_arr)
456 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
458 fails. Added test wether graph_arr is already set in phi_merge.
460 Added tons of comments to explain Phi generation -- the stack hack.
464 irdump now outputs the pnc_number for projs of Cmp nodes.
465 Added routine to access the pnc_string.
468 Added result "dataseg" to Start node, with all it's consequences.
469 We need this for the compiler lab to translate global variables.
470 Now they can be selected from the data segment, and the start
471 node supplies the pointer to this segment.
472 These changes are guarded by preprocessor variable UEBPRAKT in
474 Changed files: common.h, irgraph.ch, irnode.h
478 removed array_example, added array-heap_example and
479 array-stack_example and adapted makefile.
481 Entity now also allows type_method as owner. This is to model
482 the stack, Sel nodes that select a variable (array) from the
483 stack need an entity that describes what they select. This
484 entity gets the type of the enclosing procedure as owner.
487 Added future access routines in `irmode.[ch]' as comments,
488 that all attributes in the ir_mode struct can be user-defined.
489 Added access routines in `irgraph.[ch]', for all attributes.
492 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
493 Added `irvrfy.[ch]' to MAKEFILE
494 Added "inline" before access routines in `irnode.h', that no warnings
495 will appear furthermore.
496 Added three more access routines in `irmode.[ch]':
497 struct tarval *get_min_of_mode (ir_mode *mode);
498 struct tarval *get_max_of_mode (ir_mode *mode);
499 struct tarval *get_null_of_mode (ir_mode *mode);
502 Adaption of Alloc, Sel and Entity to new types:
503 Changed array_example to use type instead of type_class.
504 Changed entity to allow all types as owner instead of
505 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
506 Changed alloc to allow allocating variables of any type instead of
508 Adapted all testprograms to avoid warnings.
510 removed dtest ... from Makefile, replaced by test.
513 Added "run" to makefile, added array_example to makefile of testprograms.
514 Corrected testprogram array
515 Changed assert in irvrfy -- Conv.