2 Fixed some testprograms to new semantics of firm.
3 Added testprogram global_var_example.
6 Fixed a bug in irop: Size of Sel node was (entity *) instead of
7 sel_attr. Therefor array_heap_example aborted.
8 Also found a missing break in typewalk, case iro_Alloc.
11 Added a construct that unifies all things constructed by the lib
12 and needed to represent a program: irprog.ch
13 This contains a list of all types as well as a list of all ir graphs.
14 I added calls to initialization and constructors so that this is constructed
15 automatically. I changed firm.c, type.c, irgraph.c and ev. more.
18 Removed strange block struct and constructor from ircons.c.
19 Moved routines for automatic Phi construction from irgmod to ircons.
20 Sorted functions in ircons to reflect the different interfaces.
21 Renamed dataseg to globals to be conformant with documentation.
22 Added conservation of current_ir_graph to irgopt.c.
23 Corrected bug in Makefiles: wrong path to directory ./inlude.
26 Commented a whole bunch of stuff, e.g. in ircons.h (Procedure to construct)
27 We should change the naming of the Block constructor.
30 Removed acces routines to attr. "closed" of Block.
33 Removed second implementation of new_ir_node from ircons and some comments
34 concerned with the now resolved cyclic dependency.
35 Fixed some bugs in irgopt so that the compilation goes through.
38 Removed some files from the archive, after dependencies and usage are
40 - 'ir/common/strerror.c'
42 - 'ir/ident/xx_ident.h'
43 an empty file, only included by 'ident.c', but never used;
46 contained a numeric counter;
47 The files were included in 'tv.h' and implemented in 'tarval.c'. The two
48 functions 'tarval_label' and 'tarval_forall_labeled' were also never used.
49 In the old fiasco compiler, the label thing is used in the backend.
50 Search for 'babil' to see the usage.
51 Also removed the two attributes 'lab' and 'used' in the tarval struct.
54 Moved the 'new_ir_node' constructor from ircons.* to irnode.* and fixed
55 afterwards some recursive includes, so libfirm works again.
58 Did a lot of changes, which I never commented until now (4.7.00).
60 * Added new result to Start node: Pointer to global data segment.
61 * Extended Semantics of SymConst node to represent information for the
63 * Added arithmeitc nodes (Shrs, Minus ...)
65 Rearranged the directory structure and adjusted the makefiles.
66 The directories contain:
67 ir: everything for the intermediate representation (better: src?)
68 /ir: the ir itself, and standard optimizations.
69 /tv: the target value module
70 /tr: the type and entity representation
71 /adt: abstract data types
72 /common: stuff needed by all other dirs
73 /debug: debugging Unterstuetzung
75 include: external files needen as includes
76 testprograms: examples to test the lib.
78 The makefiles generate files with extension .d that contain the dependencies
82 Added access routine to attribute link of irnode in irnode.ch.
83 Added get_negated_pnc to irnode.ch.
86 Added an iterator in iropt that calls the optimization for all nodes.
87 Copied optimize's code to optimize_in_place and removed deallocations.
88 Made set_irn_n in irnode.c public.
89 Added Bad as possible cfg predecessor in is_cfg_node in irnode. Bads
90 are allowded as cfgpredecessors to represent dead "incoming" control
92 Added call to optimize_in_place in mature_block. Is this useful?
95 Changed tests from comparing enums to comparing pointers. This is more
96 efficient (is it?) and reads better. e.g., instead get_irn_opcode == irm_And
97 now get_irn_op == op_And
100 Compared equivalent_value to the one in the original fiasco.
101 Added a test in case iro_And.
102 Changed implementation of iro_Div. Added routines turn_into_tuple
103 and set_irn_op() in irgmod/irnode.
104 Changed implementation of irn_Minus (although commented out).
105 Now all features are ported or documented for this function.
108 Compared computed_value to the one in the original fiasco.
109 Added one further optimization in case iro_Proj.
110 Now all features are ported for this function.
113 Updated comments in ircons.h.
116 Changed possible attribute of SymConst from type_class to type.
117 SymConst size could be the size of a union, array, ...
120 Implemented two dumpers that dump type information and a graph with
124 Implemented a walker that walks over the type informatin that can
125 be reached from an ir node.
126 Added routine get_kind(firm_thing) in common.h, Added file common.c.
127 Added files typewalk.ch and routines to irdump.
128 Added visited flags and initialization to entity and all types.
129 Implemented some of the missing access routines in type.c, entity.c
132 Implemented constructor for ShrA.
134 Added routine that iterates the graph and verifies all nodes.
135 Added Phi_in stack to irgraph. This allows to build several
136 graphs at once with respect to the Phi building algorithm.
139 Implemented dumper for control flow graphs. Changed irr_* and dead_*
140 examples to dump cfg.
143 Implemeted explicit stack for the Phi node generation. See ircons.c.
146 added include string.h in xoprintf.c to get rid of warning:
147 implicit declaration of function `bcopy'
151 Added more cases to equivalent_node.
152 Portet gigo, identify, identify_remember.
153 The set/pset library can not handle several hash tables. At least,
154 compiling several procedures causes a segmentation fault during access
155 to the tables (oo_program_example). Need to work on this.
156 transform_node still needs to be ported.
157 some cases in equivalent_node are still missing.
159 To get it running I had to remove the call to optimize in the
160 new_Block constructor. It can only be called when the Block is
164 Introduced a walker that walks only over blocks. The asserts
165 might be too strict for general graphs, but so far it works.
166 For the walker I introduced an additional flag in the attribute
167 of Blocks, an a global block_visited variable (in irgraph.h).
168 Furhter I modularized dump_ir_node, implemented dump_ir_graph
169 with the existing walker, and added new routines to output a
171 Changed the examples to call the new dump routine.
174 Added auxiliary routines to ir_node:
175 skip_Proj, skip_Nop, is_Bad ...
176 Changed access routines to use get_irn_n instead of explicit array
177 accesses. Changed get_irn_n to call skip_Id, so that the access routines
178 never will return Id nodes.
181 Repaired Phi computation, changed new_r_Phi_in. Added test program to
182 test irregular controlflow.
185 Updated optimizations computed_value and equivalent_node to new
186 Syntax of FIRM nodes and to use access routines.
187 Added routine skip_Proj in irnode.
189 2.2.2000 Boris Boesler
190 Added files irgwalk.c and irgwalk.h. They include a simple function to
191 traverse an ir graph and execute pre and post functions to current node.
194 Added arrays with parameters and result types to the type informtion
195 of a procedure. This is needed for the code generation. Also added
196 the corresponding access routines.
197 Adapted the testprograms.
200 Found a bug / complication: the call order
201 get_value (makes Phi0, put's it into graph_arr)
202 set_value (overwrites Phi0 in graph_arr)
203 mature_block (upgrades Phi0, puts it again into graph_arr, overwriting
205 fails. Added test wether graph_arr is already set in phi_merge.
207 Added tons of comments to explain Phi generation -- the stack hack.
211 irdump now outputs the pnc_number for projs of Cmp nodes.
212 Added routine to access the pnc_string.
215 Added result "dataseg" to Start node, with all it's consequences.
216 We need this for the compiler lab to translate global variables.
217 Now they can be selected from the data segment, and the start
218 node supplies the pointer to this segment.
219 These changes are guarded by preprocessor variable UEBPRAKT in
221 Changed files: common.h, irgraph.ch, irnode.h
225 removed array_example, added array-heap_example and
226 array-stack_example and adapted makefile.
228 Entity now also allows type_method as owner. This is to model
229 the stack, Sel nodes that select a variable (array) from the
230 stack need an entity that describes what they select. This
231 entity gets the type of the enclosing procedure as owner.
234 Added future access routines in `irmode.[ch]' as comments,
235 that all attributes in the ir_mode struct can be user-defined.
236 Added access routines in `irgraph.[ch]', for all attributes.
239 Moved verify routines from `irgraph.[ch]' to own files `irvrfy.[ch]'.
240 Added `irvrfy.[ch]' to MAKEFILE
241 Added "inline" before access routines in `irnode.h', that no warnings
242 will appear furthermore.
243 Added three more access routines in `irmode.[ch]':
244 struct tarval *get_min_of_mode (ir_mode *mode);
245 struct tarval *get_max_of_mode (ir_mode *mode);
246 struct tarval *get_null_of_mode (ir_mode *mode);
249 Adaption of Alloc, Sel and Entity to new types:
250 Changed array_example to use type instead of type_class.
251 Changed entity to allow all types as owner instead of
252 only type_class. Added routine "assert_legal_owner_of_ent(type* type)".
253 Changed alloc to allow allocating variables of any type instead of
255 Adapted all testprograms to avoid warnings.
257 removed dtest ... from Makefile, replaced by test.
260 Added "run" to makefile, added array_example to makefile of testprograms.
261 Corrected testprogram array
262 Changed assert in irvrfy -- Conv.