OSDN Git Service

3c19575369a478c7d2a654546c80e4404c9b07a3
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-01-19  Richard Guenther  <rguenther@suse.de>
2
3         * builtins.c (expand_builtin_cexpi): Fall back to expanding
4         via cexp in case sincos is not available.
5
6 2007-01-19  Richard Guenther  <rguenther@suse.de>
7
8         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
9         * defaults.h (TARGET_HAS_SINCOS): Default to off.
10         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
11         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
12         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
13         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
14         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
15         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
16
17 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
18
19         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
20         New insn patterns.
21         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
22         (atan2xf3): Directly generate RTL pattern.
23         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
24         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
25         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
26         disable patterns for SSE math.
27         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
28         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
29         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
30         disable patterns for SSE math.
31
32 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
33
34         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
35
36 2007-01-18  Roger Sayle  <roger@eyesopen.com>
37
38         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
39         VIEW_CONVERT_EXPR to the same type as it's operand.
40
41 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
42
43         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
44         configured for long double 128.
45
46 2007-01-18  Mike Stump  <mrs@apple.com>
47
48         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
49         error.
50
51 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
52
53         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
54         type correct.
55
56 2007-01-18  Jan Hubicka  <jh@suse.cz>
57
58         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
59         greater than NUM_VOP_FREE_BUCKETS.
60
61 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
62
63         * tree-ssa-structalias.c: Update comments.
64         (ptabitmap_obstack): Removed.
65         (pta_obstack): New.
66         (oldpta_obstack): Ditto.
67         (stats): Add a few members.
68         (struct variable_info): Remove node, complex, address_taken, and
69         indirect_target members. Add oldsolution member.
70         (new_var_info): Do not initialize removed members.
71         (constraint_expr_type): Remove INCLUDES.
72         (constraint_graph): Add size, implicit_preds, rep,
73         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
74         (FIRST_REF_NODE): New macro.
75         (LAST_REF_NODE): Ditto.
76         (FIRST_ADDR_NODE): Ditto.
77         (find): New function.
78         (unite): Ditto.
79         (dump_constraint): Do not handle INCLUDES.
80         (insert_into_complex): Do not insert duplicate constraints.
81         (condense_varmap_nodes): Renamed and rewritten into ...
82         (merge_node_constraints): This. Also fix bug in handling of
83         offseted copy constraints.
84         (clear_edges_for_node): No longer need to deal with preds at all,
85         or removing associated preds/succs. 
86         (merge_graph_nodes): Deal with indirect_cycles.
87         Don't deal with predecessors.
88         (add_implicit_graph_edge): New function.
89         (add_pred_graph_edge): Ditto.
90         (add_graph_edge): Don't deal with predecessors.
91         (build_constraint_graph): Removed.
92         (build_pred_graph): New function.
93         (build_succ_graph): Ditto.
94         (struct scc_info): Removed in_component. Added roots, dfs, and
95         node_mapping. Remove visited_index, unification_queue.
96         (scc_visit): Deal with union-find we do now.
97         Deal with cycles with REF nodes.
98         (collapse_nodes): Renamed and rewritten to ...
99         (unify_nodes): This.
100         (process_unification_queue): Removed.
101         (topo_visit): Cleanup
102         (do_da_constraint): Use find.
103         (do_sd_constraint): Ditto.
104         (do_ds_constraint): Ditto.
105         (do_complex_constraint): Ditto.
106         (init_scc_info): Update for removed and added members.
107         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
108         (find_indirect_cycles): This.
109         (equivalence_class): New variable.
110         (label_visit): New function.
111         (perform_variable_substitution): Rewritten.
112         (free_var_substitution_info): New function.
113         (find_equivalent_node): Ditto.
114         (move_complex_constraints): Ditto.
115         (eliminate_indirect_cycles): Ditto.
116         (solve_graph): Only propagate changed bits.
117         Use indirect cycle elimination.
118         Use find.
119         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
120         (tree_id_eq): Renamed to ...
121         (tree_vi_eq): This. Update for member change
122         (insert_id_for_tree): Renamed and rewritten to ...
123         (insert_vi_for_tree): This.
124         (lookup_id_for_tree): Renamed and rewritten to ...
125         (lookup_vi_for_tree): This.
126         (get_id_for_tree): Renamed and rewritten to ...
127         (get_vi_for_tree): Ditto.
128         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
129         (process_constraint): Don't handle INCLUDES.
130         Remove special ADDRESSOF case.
131         (find_func_aliases): Rewrite to use vi functions instead of id
132         ones.
133         (create_function_info_for): Ditto.
134         (create_variable_info_for): Ditto.
135         (intra_create_variable_infos): Ditto.
136         (merge_smts_into): Ditto.
137         (find_what_p_points_to): Ditto.
138         (init_base_vars): Ditto.
139         (init_alias_vars): Ditto.
140         (remove_preds_and_fake_succs): New function.
141         (dump_sa_points_to_info): Dump new stats.
142         (dump_solution_for_var): Use find.
143         (set_used_smts): Fix formatting.
144         (compute_points_to_sets): Updated for new functions.
145         (ipa_pta_execute): Ditto.
146
147 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
148             Richard Sandiford  <richard@codesourcery.com>
149
150         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
151         parallels.
152         * calls.c (expand_call): If the return value is a PARALLEL,
153         extract its first member.
154         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
155         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
156         use FUNCTION_EXTRA_EPILOGUE.
157         (m68k_function_value): Return a PARALLEL if the return value
158         is of a pointer type.
159         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
160         (FUNCTION_EXTRA_EPILOGUE): Remove.
161         * config/m68k/m68k.md (D0_REG): New constant.
162
163 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
164
165         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
166         output a NOP for empty epilogues.
167
168 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
169
170         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
171         before function.  Extend register save check to include all
172         registers, not just integer ones.
173
174 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
175
176         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
177         (m68k_use_return_insn): ...this.
178         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
179         * config/m68k/m68k.c (use_return_insn): Rename to...
180         (m68k_use_return_insn): ...this.
181         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
182         of USE_RETURN_INSN.
183
184 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
185
186         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
187         nearest even, fix denormal rounding overflow.
188
189 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
190
191         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
192         f%$smove and f%$move.
193         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
194         (extendsfdf2_cf): Use fdmove instead of f%&move.
195         (truncdfsf2_cf): Use fsmove instead of f%$smove.
196         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
197
198 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
199
200         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
201         GPR<-GPR moves.
202
203 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
204
205         * real.h (real_format): Add a canonical_nan_lsbs_set field.
206         (coldfire_single_format): Declare.
207         (coldfire_double_format): Likewise.
208         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
209         of qnan_msb_set to determine the lower bits of a canonical
210         NaN significand.
211         (encode_ieee_double): Likewise.
212         (encode_ieee_quad): Likewise.
213         (ieee_single_format): Initialize canonical_nan_lsbs_set.
214         (mips_single_format): Likewise.
215         (ieee_double_format): Likewise.
216         (mips_double_format): Likewise.
217         (ieee_extended_motorola_format): Likewise.
218         (ieee_extended_intel_96_format): Likewise.
219         (ieee_extended_intel_128_format): Likewise.
220         (ieee_extended_intel_96_round_53_format): Likewise.
221         (ibm_extended_format): Likewise.
222         (mips_extended_format): Likewise.
223         (ieee_quad_format): Likewise.
224         (mips_quad_format): Likewise.
225         (vax_f_format): Likewise.
226         (vax_d_format): Likewise.
227         (vax_g_format): Likewise.
228         (i370_single_format): Likewise.
229         (i370_double_format): Likewise.
230         (decimal_single_format): Likewise.
231         (decimal_double_format): Likewise.
232         (decimal_quad_format): Likewise.
233         (c4x_single_format): Likewise.
234         (c4x_extended_format): Likewise.
235         (real_internal_format): Likewise.
236         (coldfire_single_format): New real_format.
237         (coldfire_double_format): Likewise.
238         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
239         canonical_nan_lsbs_set.
240         (pdp11_d_format): Likewise.
241         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
242         if TARGET_COLDFIRE_CPU.
243
244 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
245
246         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
247         (output_call, m68k_legitimize_call_address): Declare.
248         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
249         parenthesees.  Add support for a 'W' constraint.
250         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
251         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
252         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
253         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
254         variables.
255         (override_options): Initialize them.  Do not set flag_no_function_cse
256         for TARGET_ID_SHARED_LIBRARY.
257         (m68k_output_pic_call): Delete.
258         (m68k_legitimize_call_address): New function.
259         (print_operand): Remove the %o prefix.  Handle the %p prefix.
260         (output_call): New function.
261         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
262         target address from the GOT if symbolic jumps are not allowed.
263         * config/m68k/m68k.md (call, general_operand): Do not set
264         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
265         Merge separate flag_pic and !flag_pic define_insns into...
266         (*call, *call_value): ...these new patterns.  Match the address
267         rather than the containing MEM and require it to be a call_operand.
268         Use output_call to generate the asm template.
269         * config/m68k/predicates.md (const_call_operand): New predicate.
270         (call_operand): Likewise.
271
272 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
273
274         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
275         (STACK_GROWS_DOWNWARD): Define to 1.
276         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
277         comments, emphasizing that these values are only defaults.
278         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
279         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
280
281 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
282
283         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
284
285 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
286             Richard Sandiford  <richard@codesourcery.com>
287
288         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
289         (FP_REGNO_P): New macros.
290         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
291         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
292         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
293         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
294
295 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
296
297         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
298         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
299         to $tmake_file.
300         * config/m68k/t-floatlib: New file, extracting common code from...
301         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
302         * config/m68k/t-uclinux: Here.
303         * config/m68k/fpgnulib.c: Do not compile extendeed precision
304         routines on ColdFire targets.
305
306 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
307
308         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
309         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
310         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
311         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
312         explicitly set MOTOROLA to 1.
313         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
314
315 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
316             Nathan Sidwell  <nathan@codesourcery.com>
317
318         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
319         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
320         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
321         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
322         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
323         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
324         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
325         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
326         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
327         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
328         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
329         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
330         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
331         (DBX_REGISTER_NUMBER): Delete.
332         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
333         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
334         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
335
336 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
337
338         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
339         * config/flat.h: New file.
340         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
341         OBJECT_FORMAT_FLAT.
342         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
343         indirect references for -msep-data or -mid-shared-library.
344         Do not use PC-relative code addresses either.
345         * config/m68k/m68k.c (override_options): Restrict -fPIC error
346         to -mpcrel.
347         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
348         for shared libraries and crt1.o for executables.  Use crti.o and
349         crtbegin.o.
350         (ENDFILE_SPEC): Use crtend.o and crtn.o.
351         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
352         Do not add -elf2flt or -shared-lib-id options here.
353         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
354         Pass -shared-lib-id if -mid-shared-library, taking the library
355         identifier from -mshared-library-id if given, otherwise
356         defaulting to 0.
357         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
358         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
359         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
360         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
361         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
362         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
363         and crtend.o.
364         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
365         jump sequence for ISA A and ISA A+.
366         (PICJUMP): Likewise.
367
368 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
369
370         * config.gcc (m68k-*-uclinux*): Base the port on the common
371         and m68k GNU/Linux files rather than on the generic ELF ones.
372         * config/m68k/uclinux.h (TARGET_VERSION): Override.
373         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
374
375 2007-01-18  Julian Brown  <julian@codesourcery.com>
376             Richard Sandiford  <richard@codesourcery.com>
377
378         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
379         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
380         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
381         before redefining.
382         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
383         80 unconditionally.
384         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
385
386 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
387
388         * doc/install.texi: Document m68k-uclinuxoldabi.
389         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
390         * config/m68k/uclinux-oldabi.h: New file, copied from
391         config/m68k/uclinux.h.
392
393 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
394
395         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
396         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
397         (m32r_compute_frame_size): Use unsigned for regno.
398         (m32r_reload_lr): Use gen_frame_mem.
399         (pop): New.
400         (m32r_output_function_epilogue): Don't output the function epilogue
401         textually here.
402         (m32r_expand_epilogue): New.
403         (direct_return): Return false if the function has the interrupt
404         attribute.
405         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
406         * config/m32r/m32r.md (epilogue): New expander.
407         (return_lr, return_rte): New insns.
408         (return): Make it expander.
409         (return_normal): New expander.
410
411 2007-01-18  Josh Conner  <jconner@apple.com>
412
413         PR target/30485
414         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
415         support for UNLE, UNLT, UNGE, and UNGT.
416
417 2007-01-18  Jan Hubicka  <jh@suse.cz>
418
419         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
420         * tree-tailcall (add_virtual_phis): Likewise.
421         (optimize_tail_call): Return TODOs.
422         (execute_tail_calls): Return TODOs.
423         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
424         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
425         changed.
426         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
427         Cleanup cfg using TODO.
428         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
429         * passes.c (execute_function_todo): When cleanup did something, remove
430         unused locals.
431         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
432         (make_edges): Don't cleanup_cfg.
433
434 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
435
436         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
437         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
438         operate on the top of stack.
439         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
440         register handling of instructions that output to the second
441         stack slot.
442         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
443         (move_for_stack_reg): Special-case check for dead destination
444         stack slot for constant load of 1.0 inside UNSPEC_TAN.
445
446         * config/i386/i386.md (UNSPEC_TAN): New constant.
447         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
448         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
449         to correctly model move of constant 1.0 to top stack slot.
450         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
451         (unnamed peephole2 pattern): Remove corresponding peephole2
452         pattern that optimizes tan insn and loading of constant 1.0.
453         (tanxf2): Use fptanxf4_i387.
454         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
455         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
456         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
457         disable patterns for SSE math.
458         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
459         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
460
461 2007-01-18  Dirk Mueller  <dmueller@suse.de>
462             Richard Guenther <rguenther@suse.de>
463
464         PR diagnostic/8268
465         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
466         * common.opt (Warray-bounds): Add new warning option.
467         * c-opts.c (c_common_handle_option): Define -Warray-bounds
468         if -Wall is given.
469         * Makefile.in: make tree-vrp.o depend on toplev.h
470         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
471         is enabled.
472         (check_array_refs, check_array_bounds, check_array_ref): New.
473
474 2007-01-18  Jan Hubicka  <jh@suse.cz>
475
476         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
477         (execute_ssa_ccp): Return flags conditionally.
478         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
479         changed.
480         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
481
482 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
483
484         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
485         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
486
487         * Makefile.c (GTFILES): Add cfglayout.h.
488         * gengtype.c (open_base_files): Likewise.
489         * cfglayout.c (cfg_layout_function_footer,
490         cfg_layout_function_header) Reindent to make gengtype happy.
491         * cfglayout.h (cfg_layout_function_footer,
492         cfg_layout_function_header): Add GTY(()) marker.
493
494         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
495
496 2007-01-18  Ben Elliston  <bje@au.ibm.com>
497
498         * genautomata.c (write_automata): Include xstrerror output in the
499         error message if writing the DFA description file fails.
500
501 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
502
503         * config/mips/mips-protos.h (mips_output_external): Make it
504         return void.
505         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
506         * config/mips/mips.c (irix_output_external_libcall): Likewise.
507         (extern_list): Likewise.
508         (extern_head): Likewise.
509         (TARGET_ASM_FILE_END): Likewise.
510         (mips_file_end): Likewise.
511         (mips_output_external): Rewritten.
512
513 2007-01-18  Ben Elliston  <bje@au.ibm.com>
514
515         * genpreds.c (write_insn_preds_c): Only write out the function
516         body for regclass_for_constraint if we have register constraints.
517
518 2007-01-17  Tom Tromey  <tromey@redhat.com>
519
520         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
521         * doc/install.texi (Testing): Use sourceware.org.
522         (Binaries): Likewise.
523         (Specific): Likewise.
524         * doc/contrib.texi (Contributors): Use sourceware.org.
525
526 2007-01-17  Anatoly Sokolov <aesok@post.ru>
527
528         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
529         (AVR_ENHANCED): Rename to ...
530         (AVR_HAVE_MUL): ... new.
531         (avr_enhanced_p): Rename to ...
532         (avr_have_mul_p): ... new.
533         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of 
534         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
535         * config/avr/avr.c (avr_enhanced_p): Rename to ...
536         (avr_have_mul_p): ... new.
537         (base_arch_s): Rename 'enhanced' to 'have_mul'.
538         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
539         'avr_enhanced_p' and 'enhanced'.
540         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use 
541         AVR_HAVE_MUL instead of AVR_ENHANCED.
542         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
543         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3, 
544         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL 
545         instead of AVR_ENHANCED.
546         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
547         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
548         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
549
550 2007-01-17  Ian Lance Taylor  <iant@google.com>
551
552         * vec.h (VEC_reserve_exact): Define.
553         (vec_gc_p_reserve_exact): Declare.
554         (vec_gc_o_reserve_exact): Declare.
555         (vec_heap_p_reserve_exact): Declare.
556         (vec_heap_o_reserve_exact): Declare.
557         (VEC_OP (T,A,reserve_exact)): New static inline function, three
558         versions.
559         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
560         negative parameter.
561         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
562         (VEC_OP (T,A,copy)) [all versions]: Likewise.
563         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
564         * vec.c (calculate_allocation): Add exact parameter.  Change all
565         callers.
566         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
567         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
568         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
569         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
570         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
571         (vec_heap_p_reserve_exact): New function.
572         (vec_heap_o_reserve_exact): New function.
573
574 2007-01-17  Jan Hubicka  <jh@suse.cz>
575
576         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
577
578 2007-01-17  Eric Christopher  <echristo@apple.com>
579
580         * config.gcc: Support core2 processor.
581
582 2007-01-16  Jan Hubicka  <jh@suse.cz>
583
584         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
585         the names, just unlink the chain so we don't crash on dangling pointers
586         to dead SSA names.
587
588 2007-01-16  Jan Hubicka  <jh@suse.cz>
589
590         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
591         * tree-pass.h: Reorder to make IPA passes appear toegher.
592         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
593         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
594         parameters, do not call early inliner.
595         * ipa-inline.c: Update comments.  Include tree-flow.h
596         (cgraph_decide_inlining): Do not compute inlining parameters.
597         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
598         be called with function context set up.
599         (pass_ipa_inline): Remove unreachable functions before pass.
600         (cgraph_early_inlining): Simplify assuming to be called from the
601         PM as local pass.
602         (pass_early_inline): New pass.
603         (cgraph_gate_ipa_early_inlining): New gate.
604         (pass_ipa_early_inline): Turn into simple wrapper.
605         (compute_inline_parameters): New function.
606         (gate_inline_passes): New gate.
607         (pass_inline_parameters): New pass.
608         (apply_inline): Move here from tree-optimize.c
609         (pass_apply_inline): New pass.
610         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
611         transforming.
612         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
613         doing them by hand.
614         (tree_function_versioning): Do not allocate dummy struct function.
615         * tree-inline.h (optimize_inline_calls): Update prototype.
616         * tree-optimize.c (execute_fixup_cfg): Export.
617         (pass_fixup_cfg): Remove
618         (tree_rest_of_compilation): Do not apply inlines.
619         * tree-flow.h (execute_fixup_cfg): Declare.
620         * Makefile.in (gt-passes.c): New.
621         * passes.c: Include gt-passes.h
622         (init_optimization_passes): New passes.
623         (nnodes, order): New static vars.
624         (do_per_function_toporder): New function.
625         (execute_one_pass): Dump current pass here.
626         (execute_ipa_pass_list): Don't dump current pass here.
627
628 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
629
630         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
631         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
632
633 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
634
635         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
636         (strong_alias): Define.
637         (__gcc_qmul): Provide non-FMA for soft-float.
638         (__gcc_qdiv): Same.
639         (__gcc_qneg): New.
640         (__gcc_qeq): New.
641         (__gcc_qle): New.
642         (__gcc_qge): New.
643         (__gcc_qunord): New.
644         (__gcc_stoq): New.
645         (__gcc_dtoq): New.
646         (__gcc_qtos): New.
647         (__gcc_qtod): New.
648         (__gcc_qtoi): New.
649         (__gcc_qtou): New.
650         (__gcc_itoq): New.
651         (__gcc_utoq): New.
652         (fmsub): New.
653         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
654         soft-float functions.
655         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
656         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
657         about long double soft float.
658
659 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
660             Tehila Meyzels  <tehila@il.ibm.com>
661
662         * tree-vectorizer.h (is_pattern_stmt_p): New.
663         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
664         formatting (tabs instead of spaces). Cleanup and clarify setting
665         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
666         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
667         (vectorizable_type_demotion): Check that types are integral.
668         (vectorizable_type_promotion): Likewise.
669         (vectorizable_store): Fix typo.  Eliminate new-line at end of
670         comments.
671
672 2007-01-16  Jan Hubicka  <jh@suse.cz>
673
674         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
675         cgraph edges too.
676
677 2007-01-15  Eric Christopher  <echristo@apple.com>
678
679         * ifcvt.c: Include vec.h, vecprim.h.
680         (check_cond_move_block): New argument regs.
681         Reorganize. Add registers used to regs.
682         (cond_move_process_if_block): Use regs set above as
683         loop bounds.
684
685 2007-01-15  Eric Christopher  <echristo@apple.com>
686
687         * config/darwin.h: Update copyright.
688         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
689         (LINK_COMMAND_SPEC): Add -u.
690         (LINK_SPEC): Fix umbrella for above.
691
692 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
693
694         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
695         glibc CVS.
696
697 2007-01-15  Tom Tromey  <tromey@redhat.com>
698
699         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
700         * doc/install.texi (Testing): Don't mention jacks.
701         (Configuration): Document --enable-java-maintainer-mode.  Move
702         --with-java-home to libgcj-specific section.  Document
703         --with-ecj-jar.
704         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
705
706 2007-01-15  Jan Hubicka  <jh@suse.cz>
707
708         * tree-ssa-dce.c (DCE_TODOs): New.
709         (propagate_necessity): Return if something changed.
710         (eliminate_unnecessary_stmts): Likewise.
711         (perform_tree_ssa_dce): Return TODO flags when needed.
712         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
713
714 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
715
716         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
717         (fyl2x_extend<mode>xf3_i387): New insn pattern.
718         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
719         patterns using X87MODEF12 mode macro.  Extend operand 1
720         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
721         (log10<mode>2): Ditto.
722         (log2<mode>2): Ditto.
723         (log1p<mode>2): Ditto.
724         (logb<mode>2): Ditto.
725         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
726         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
727         (*fxtractxf3_i387): Rename from *fxtractxf3.
728         (fxtract_extend<mode>xf3_i387): New insn pattern.
729         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
730         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
731         and gen_fyl2xxf3_i387().
732
733 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
734
735         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
736         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
737         get_loop_exit_edges.  Changed return type to bool.
738         * cfgloop.h (fix_loop_placement): Declaration removed.
739
740 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
741
742         * param.h (MIN_VECT_LOOP_BOUND): New.
743         * params.def (MIN_VECT_LOOP_BOUND): New.
744         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
745         argument - minimum threshold for number of iterations.
746         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
747         argument to declaration.
748         * tree-vect-analyze.c (vect_analyze_operations): Check value of
749         MIN_VECT_LOOP_BOUND.
750         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
751         slpeel_tree_peel_loop_to_edge with additional argument.
752         (vect_do_peeling_for_alignment): Likewise.
753         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
754
755 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
756
757         PR target/30413
758         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
759         operands of size 1.
760
761 2007-01-14  Jan Hubicka  <jh@suse.cz>
762
763         * tree-dfa.c (remove_referenced_var): New function.
764         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
765         prune referenced vars list too.
766         * tree-flow.h (remove_referenced_var): Declare.
767
768 2007-01-14  Jan Hubicka  <jh@suse.cz>
769
770         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
771         separately.
772         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
773         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
774         * except.h (verify_eh_throw_table_statements): Kill prototype.
775         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
776         into call exrepssions.
777         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
778         eh trhow verifier.
779         * tree-cfg.c: Include pointer-set.h.
780         (verify_node_sharing): Work on pointer set.
781         (verify_eh_throw_stmt_node): New.
782         (verify_stmts): Use pointers sets, verify throw_stmt.
783
784 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
785
786         * ipa-reference.c (analyze_function): Consider also addresses taken
787         in phi nodes.
788
789 2007-01-12  Roger Sayle  <roger@eyesopen.com>
790
791         * c-typeck.c (null_pointer_constant_p): Replace use of
792         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
793         (build_c_cast): Likewise.
794
795 2007-01-12  Roger Sayle  <roger@eyesopen.com>
796
797         * tree.h (force_fit_type_double): Remove unused final argument.
798         * c-common.c (constant_expression_warning): Replace use of
799         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
800         (convert_and_check): Likewise.
801         (shorten_compare): Update call to force_fit_type_double.
802         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
803         * convert.c (convert_to_pointer): Update call to
804         force_fit_type_double.
805         * fold-const.c (force_fit_type_double): Remove overflowed_const
806         argument.
807         (int_const_binop, fold_convert_const_int_from_int,
808         fold_convert_const_int_from_real, fold_div_compare,
809         fold_sign_changed_comparison, fold_unary, fold_negate_const,
810         fold_abs_const, fold_not_const): Remove the final argument from
811         calls to force_fit_type_double.
812
813 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
814
815         * configure.ac: Set insn to "nop" for spu-*-* also.
816         * configure: Regenerate.
817
818 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
819
820         * builtins.def : Add BUILT_IN_FREE.
821
822 2007-01-12  Jan Hubicka  <jh@suse.cz>
823
824         PR tree-optimization/30443
825         * tree-inline.c (tree_function_versioning): Do not optimize when
826         cloning for inlining.
827
828 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
829
830         * doc/loop.texi: Document recording of loop exits.
831         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
832         (update_single_exits_after_duplication,
833         update_single_exit_for_duplicated_loop,
834         update_single_exit_for_duplicated_loops): Removed.
835         (duplicate_loop_to_header_edge): Do not call
836         update_single_exits_after_duplication and
837         update_single_exit_for_duplicated_loops.
838         (loop_version): Do not update single_exit information.
839         (fix_loop_structure): Use record_loop_exits instead of
840         mark_single_exit_loops.
841         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
842         the lists of loop exits.
843         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
844         split_edge, merge_blocks): Update the lists of loop exits.
845         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
846         loop_optimizer_init.
847         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
848         of mark_single_exit_loops.
849         (loop_optimizer_finalize): Call release_recorded_exits.
850         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
851         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
852         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
853         update single exit information.
854         * lambda-code.c (perfect_nestify): Ditto.
855         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
856         (mark_single_exit_loops): Removed.
857         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
858         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
859         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
860         functions.
861         (get_loop_exit_edges, single_exit): Use recorded exit lists.
862         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
863         (verify_loop_structure): Verify consistency of the exit lists.
864         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
865         (set_single_exit): Removed.
866         * cfgloop.h (struct loop_exit): New function.
867         (struct loop): single_exit_ field replaced by exits field.
868         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
869         (struct loops): Added exits hash.
870         (mark_single_exit_loops, set_single_exit): Declaration removed.
871         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
872
873 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
874
875         * doc/invoke.texi: Avoid use of @headitem.
876
877 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
878
879         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
880
881 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
882
883         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
884         list of targets.
885
886 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
887             Richard Sandiford  <richard@codesourcery.com>
888
889         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
890         * config/m68k/m68k.opt: Resort options.
891         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
892
893 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
894             Richard Sandiford  <richard@codesourcery.com>
895
896         * doc/invoke.texi: Document the macros that are defined by
897         m68k's -mtune and -mhard-float options.
898         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
899         (m68k_cpp_cpu_family): Likewise.
900         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
901         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
902         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
903         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
904         (m68k_cpp_cpu_family): Likewise.
905
906 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
907
908         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
909         macros besides mc68000 as tuning macros.  Use a switch statement
910         to set them and mcpu32.
911
912 2007-01-12  Julian Brown  <julian@codesourcery.com>
913
914         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
915         (TARGET_68040_ONLY): Rename to...
916         (TARGET_68040): ...this.
917         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
918         * config/m68k/m68k.md: Likewise.
919
920 2007-01-12  Julian Brown  <julian@codesourcery.com>
921             Nathan Sidwell  <nathan@codesourcery.com>
922             Richard Sandiford  <richard@codesourcery.com>
923
924         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
925         the -mcpu= argument associated with the --with-cpu setting.
926         Define M68K_DEFAULT_TUNE to the default -mtune= option,
927         if different from the one implied by the -mcpu setting.
928         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
929         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
930         * doc/invoke.texi: Mention ColdFire in the introduction to the
931         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
932         -mno-div and -mhard-float options.  Make -m68881 a synonym for
933         -mhard-float.  Document the previously-undocumented -m5206e,
934         -m528x, -m5307 and -m5407 options.  Tweak the existing option
935         documentation for consistency.
936         * doc/install.texi: Mention new --with-cpu arguments.
937         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
938         default CPU if neither -mcpu nor -march are specified.
939         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
940         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
941         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
942         using m68k_tune to decide between families that implement the
943         same ISA.  Use m68k_tune to set __mcfv4e__.
944         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
945         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
946         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
947         (FL_ISA_C, FL_ISA_MMU): New macros.
948         (MASK_COLDFIRE): Delete.
949         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
950         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
951         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
952         (TARGET_HARD_FLOAT): Do not define here.
953         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
954         (TUNE_68000): New macro.
955         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
956         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
957         (TUNE_CFV2): Redefine in terms of m68k_tune.
958         (uarch_type, target_device, fpu_type): New enums.
959         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
960         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
961         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
962         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
963         (FL_FOR_isa_c): New macros.
964         (m68k_isa): New enum.
965         (m68k_target_selection): New structure.
966         (all_devices, all_isas, all_microarchs): New tables.
967         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
968         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
969         (MASK_ALL_CPU_BITS): Delete.
970         (m68k_find_selection): New function.
971         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
972         Map the legacy target options to a combination of the new ones.
973         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
974         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
975         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
976         settings.
977         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
978         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
979         properties.
980         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
981         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
982         * config/m68k/m68k-devices.def: New file.
983
984 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
985             Nathan Sidwell  <nathan@codesourcery.com>
986
987         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
988         (SUBTARGET_EXTRA_SPECS): New macros.
989         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
990         use %(asm_cpu_spec) instead.
991         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
992         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
993         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
994         (EXTRA_SPECS): Rename to...
995         (SUBTARGET_EXTRA_SPECS): ...this.
996
997 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
998             Richard Sandiford  <richard@codesourcery.com>
999             Julian Brown  <julian@codesourcery.com>
1000
1001         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1002         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1003         to the configuration's default CPU.
1004         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1005         Remove default masks.
1006         (m680[012]0-*-*): Set the default with_cpu to the first part of
1007         the target name.
1008         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1009         (m68k*-*-linux): Extend the --with-cpu handling to...
1010         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
1011         and m68010.  Don't set target_cpu_default2.
1012         * doc/install.texi: Document --with-cpu for m68k.
1013         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1014         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1015         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1016         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1017         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1018         (ASM_SPEC): Remove use of %(asm_cpu_default).
1019         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1020         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1021         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
1022         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1023         __HAVE_FPU__ if TARGET_HARD_FLOAT.
1024         (TARGET_DEFAULT): Delete.
1025         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1026         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1027         (CPP_CPU_SPEC): Delete.
1028         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1029         the appropriate string.
1030         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1031         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1032         (ASM_SPEC): Don't use %(asm_default_spec).
1033         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1034         TARGET_DEFAULT and add MASK_68881.
1035         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1036
1037 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1038
1039         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1040         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1041         MASK_68010 alongside MASK_68020.
1042         * doc/invoke.texi: Document -m68010.
1043         * config/m68k/m68k.opt (m68010): New.
1044         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1045         if TUNE_68010.
1046         (TUNE_68010): New macro.
1047         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1048         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1049         (M68K_CPU_m68332): Add MASK_68010.
1050         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1051         fallback definition.
1052         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1053         defines.
1054         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1055         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
1056         to all entries that use MASK_68020.
1057         (output_move_simode_const, output_move_himode, output_move_qimode)
1058         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1059         instead of TARGET_68020 to select clr behavior.  Remove comment
1060         about there being no TARGET_68010.
1061         * config/m68k/m68k.md: Likewise throughout.
1062
1063 2007-01-12  Julian Brown  <julian@codesourcery.com>
1064
1065         * config/m68k/m68k.h (TARGET_ISAB): New macro.
1066         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1067         * config/m68k/m68k.md: Likewise.
1068
1069 2007-01-12  Julian Brown  <julian@codesourcery.com>
1070
1071         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1072         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1073
1074 2007-01-12  Julian Brown  <julian@codesourcery.com>
1075
1076         * config/m68k/m68k.h (TUNE_68040_60): New macro.
1077         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1078         * config/m68k/m68k.md: Likewise.
1079
1080 2007-01-12  Julian Brown  <julian@codesourcery.com>
1081             Richard Sandiford  <richard@codesourcery.com>
1082
1083         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1084         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
1085         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
1086         TARGET_CPU32.
1087         (TARGET_CPU32): Rename to...
1088         (TUNE_CPU32): ...this.
1089         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
1090         (TUNE_CFV2): New macros.
1091         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
1092         remove conditions that are implied by TARGET_68020.
1093         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
1094         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
1095         (m68k_output_function_epilogue): Likewise.
1096         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
1097         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
1098         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
1099         non-68000 timings.  Refactor multiplication and division costs.
1100         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
1101         TUNE_CPU32 instead of TARGET_CPU32.
1102         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
1103         and TUNE_68060 instead of TARGET_68060.
1104         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
1105         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
1106         TARGET_CPU32.
1107         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
1108         && !TARGET_COLDFIRE" to choose between moveq and clr.
1109         Likewise in the unnamed movsf pattern.
1110         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
1111         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
1112         ashiftrt pattern.
1113
1114 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1115
1116         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
1117         of tabbing before backslashes.
1118
1119 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
1120
1121         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
1122         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
1123         (ASM_OUTPUT_INTERNAL_LABEL): Define.
1124
1125 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
1126
1127         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
1128         find_interesting_uses_cond.
1129         (find_interesting_uses_cond): Use extract_cond_operands.
1130         (rewrite_use_compare): Use extract_cond_operands and
1131         force_gimple_operand_bsi.  Do not call update_stmt.
1132         (determine_use_iv_cost_condition): Use extract_cond_operands.
1133         Return cheaper of using original bound and changing the exit bound.
1134
1135 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
1136
1137         PR tree-optimization/29516
1138         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
1139         most_expensive_mult_to_index, addr_to_parts,
1140         create_mem_ref, maybe_fold_tmr): Make the type of
1141         fields of TARGET_MEM_REF sizetype.
1142         (move_fixed_address_to_symbol, move_pointer_to_base):
1143         New functions.
1144         * tree.def (TARGET_MEM_REF): Add comment on types of
1145         the operands.
1146
1147 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
1148
1149         * c-common.c (vector_types_convertible_p): Treat opaque types as
1150         always convertible if they have the same size, but not otherwise.
1151
1152 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
1153
1154         * ifcvt.c (struct noce_if_info): Add comments to the fields.
1155         Remove the b_unconditional field.
1156         (noce_try_sign_mask): Do not look at b_unconditional.
1157         (noce_process_if_block): Do not use merge_if_blocks.  Update
1158         the CFG here.  Do not set b_unconditional.
1159         (cond_move_process_if_block): Likewise.
1160         (find_cond_trap): Likewise.
1161         (check_cond_move_block): Require simple jump insns at the end
1162         of the basic block.
1163
1164 2007-01-11  Jan Hubicka  <jh@suse.cz>
1165
1166         PR tree-optimization/1046
1167         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
1168         when alias info is not ready.
1169         (pass_tail_recursion): Do not require aliasing.
1170         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
1171         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
1172         * tree-ssa-copy.c (pass_copy_prop): Likewise.
1173         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
1174         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
1175         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
1176         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
1177         before inlining.
1178         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
1179         When aliasing is not build, mark statement as volatile.
1180
1181 2007-01-11  Tom Tromey  <tromey@redhat.com>
1182
1183         PR preprocessor/15185, PR preprocessor/20989:
1184         * doc/cppopts.texi <-MT>: Update description of algorithm for
1185         computing default target.
1186         <-M, -MD>: Reword "basename" text.
1187
1188 2007-01-11  Roger Sayle  <roger@eyesopen.com>
1189
1190         * builtins.c (expand_builtin_pow, expand_builtin_powi,
1191         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
1192         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
1193         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
1194         fold_builtin_bswap, real_constp, fold_builtin_pow,
1195         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
1196         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
1197         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1198         * convert.c (convert_to_pointer): Likewise.
1199         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
1200         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
1201         const_binop, fold_convert_const_int_from_int,
1202         fold_convert_const_int_from_real,
1203         fold_convert_const_real_from_real, sign_bit_p,
1204         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
1205         fold_sign_changed_comparison, fold_unary, fold_comparison,
1206         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
1207         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
1208         * print-tree.c (print_node_brief, print_node): Likewise.
1209         * stor-layout.c (place_field, layout_type): Likewise.
1210         * tree-chrec.c (keep_cast): Likewise.
1211         * tree.c (build_vector, build_real, build_real_from_int_cst,
1212         build_complex): Likewise.
1213
1214 2007-01-11  Roger Sayle  <roger@eyesopen.com>
1215
1216         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
1217         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
1218
1219 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
1220
1221         * configure.ac (strict1_warn): Rename to strict_warn.
1222         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
1223         cc_set_by_configure, quoted_cc_set_by_configure,
1224         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
1225         all_boot_languages, all_stagestuff): Remove.
1226         (target_list): Remove bootstrap targets.
1227         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
1228         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
1229         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
1230         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
1231         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
1232         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
1233         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
1234         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
1235         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
1236         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
1237         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
1238         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
1239         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
1240         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
1241         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
1242         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
1243         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
1244         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
1245         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
1246         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
1247         (mostlyclean): Adjust.
1248         (clean, distclean): Don't mention bootstrap stuff.
1249         * configure: Regenerate.
1250         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
1251         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
1252         treelang/config-lang.in (stagestuff): Remove.
1253         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
1254         stageprofile, stagefeedback, stagestuff): Remove mention.
1255
1256 2007-01-11  Nick Clifton  <nickc@redhat.com>
1257
1258         * config/mcore/predicates.md (mcore_general_movesrc_operand):
1259         Accept CONSTs.
1260         (mcore_general_movdst_operand): Do not accept CONST_INTs.
1261         (mcore_arith_K_S_operand): Run the test for the S constraint not
1262         the test for the M constraint.
1263         (mcore_addsub_operand): Do not accept integer values that are
1264         larger than 32 bits.
1265         * config/mcore/mcore.md: Remove unused constraints from split.
1266         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
1267         (addsi3): Likewise.
1268         (allocate_stack): Likewise.
1269         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
1270         operands to 32 bits.
1271         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
1272         INTVAL.
1273         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
1274         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
1275         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
1276         mcore_output_bclri, mcore_output_andn, output_inline_const,
1277         mcore_output_move, mcore_output_movedouble): Likewise.
1278         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
1279         (output_inline_const): Likewise.
1280         (output_inline_const): Fix format strings used in sprintf
1281         statements.
1282         * config/mcore/mcore-protos.h: Update prototypes for changed
1283         functions in mcore.c.
1284         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
1285         HOST_WIDE_INT and not int.
1286         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
1287         CONST_OK_FOR_N): Likewise.
1288         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
1289         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
1290         an INTVAL.
1291
1292 2007-01-10  Jan Hubicka  <jh@suse.cz>
1293
1294         * tree-vrp.c (remove_range_assertions): Release defs.
1295         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
1296         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
1297
1298 2007-01-10  Paul Brook  <paul@codesourcery.com>
1299
1300         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
1301         * config/arm/arm.md (mulsi3subsi): New insn.
1302
1303 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
1304
1305         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
1306         (tree_transform_and_unroll_loop): New.
1307         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
1308         Declare.
1309
1310 2007-01-10  Robert Kennedy <jimbob@google.com>
1311
1312         * fold-const.c (fold_comparison): Fold comparisons like (x *
1313         1000 < 0) to (x < 0).
1314
1315 2007-01-10  Ian Lance Taylor  <iant@google.com>
1316
1317         * tree-pretty-print.c (dump_generic_node): Print parentheses when
1318         operands have the same priority.
1319
1320 2007-01-10  Tom Tromey  <tromey@redhat.com>
1321
1322         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
1323         (fold_binary): Likewise.
1324         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
1325         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
1326         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
1327         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
1328         * langhooks.h (struct lang_hooks): Removed field
1329         'can_use_bit_fields_p'.
1330
1331 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
1332
1333         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
1334
1335 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
1336
1337         * function.c (get_last_funcdef_no): New function.
1338         * function.h (get_last_funcdef_no): Declare.
1339         * tree-inline.c (initialize_cfun): Add initialization.
1340         (tree_function_versioning): Cleanup.
1341
1342 2007-01-10  Jan Hubicka  <jh@suse.cz>
1343
1344         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
1345         PHIs.
1346
1347 2007-01-10  Sa Liu  <saliu@de.ibm.com>
1348             Ben Elliston  <bje@au.ibm.com>
1349
1350         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
1351         for SAVE_NONLOCAL and Pmode for any other save level.
1352         * spu-protos.h (spu_restore_stack_block): Declare.
1353         * spu.md (save_stack_block): Remove.
1354         (restore_stack_block): Call spu_restore_stack_block.
1355         * spu.c (spu_restore_stack_block): New function.
1356         (spu_expand_epilogue): Remove old comment.
1357
1358 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
1359
1360         PR tree-optimization/30322
1361         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
1362         (cand_value_at): Return the value as aff_tree.
1363         (may_eliminate_iv): Convert the bound from aff_tree to tree.
1364         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
1365         aff_combination_mult): New functions.
1366         (aff_combination_add): Use aff_combination_add_cst.
1367         (aff_combination_convert): Allow conversions to a wider type.
1368         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
1369         * tree-affine.h (aff_combination_mult): Declare.
1370
1371 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
1372
1373         * doc/tm.texi: Update documentation to reflect reality of exec
1374         and start file search behaviours. Update copyright year.
1375         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
1376         header file directories.
1377
1378 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
1379
1380         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
1381         (*cosxf2): Rename to cosxf2_i387.
1382         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
1383         using X87MODEF12 mode macro. Rename patterns to
1384         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
1385         for SSE math.
1386         (*cosdf2, *cossf2): Ditto.
1387         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
1388         to match extended input operands.
1389         (sincos<mode>3): New expander.
1390         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
1391         insn patterns and corresponding splitters.
1392
1393 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
1394
1395         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
1396         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
1397
1398 2007-01-09  Nicolas Pitre  <nico@cam.org>
1399
1400         PR target/30173
1401         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
1402
1403 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
1404
1405         * target.h (struct gcc_target): New field library_rtti_comdat.
1406         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
1407         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
1408         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
1409         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1410
1411 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
1412
1413         * doc/invoke.texi (Optimize Options): Correct description of -O0.
1414
1415 2007-01-08  Richard Guenther  <rguenther@suse.de>
1416
1417         * tree.h (force_fit_type_double): Export.
1418         (force_fit_type): Remove.
1419         * fold-const.c (force_fit_type_double): New function.
1420         (force_fit_type): Remove.
1421         (int_const_binop): Use it.
1422         (fold_convert_const_int_from_int): Likewise.
1423         (fold_convert_const_int_from_real): Likewise.
1424         (fold_div_compare): Likewise.
1425         (fold_sign_changed_comparison): Likewise.
1426         (fold_unary): Likewise.
1427         (fold_negate_const): Likewise.
1428         (fold_abs_const): Likewise.
1429         (fold_not_const): Likewise.
1430         * c-common.c (shorten_compare): Use force_fit_type_double.
1431         * convert.c (convert_to_pointer): Likewise.
1432
1433 2007-01-08  Richard Guenther  <rguenther@suse.de>
1434
1435         * tree.h (build_int_cst_wide_type): Export.
1436         * tree.c (build_int_cst_wide_type): New function.
1437         (build_int_cst_wide): Fix comment.
1438         * builtins.c (fold_builtin_object_size): Use build_int_cst
1439         to build -1 or 0 of the correct type.  Use fit_double_type
1440         to check for overflow.
1441         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
1442         to build the mask.
1443         (decode_field_reference): Likewise.
1444         (all_ones_mask_p): Likewise.
1445         (native_interpret_int): Use build_int_cst_wide_type.
1446         (fold_binary): Use build_int_cst_type to build an all-ones
1447         value.
1448         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
1449
1450 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
1451
1452         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
1453
1454 2007-01-09  Ben Elliston  <bje@au.ibm.com>
1455
1456         * genautomata.c (STATS_OPTION): New option.
1457         (stats_flag): New flag.
1458         (gen_automata_option): Handle it.
1459         (initiate_automaton_gen): Ditto.
1460         (write_automata): Output statistics only if stats_flag is
1461         set. Likewise, output time statistics only if time_flag is set.
1462         * doc/md.texi (Processor pipeline description): Document new flag.
1463
1464 2007-01-08  Richard Guenther  <rguenther@suse.de>
1465
1466         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
1467         * tree.c (build_int_cst_type): Likewise.
1468         (size_in_bytes): Don't call force_fit_type on the result.
1469         (int_fits_type_p): Use fit_double_type.
1470         * fold-const.c (fit_double_type): New function.
1471         (force_fit_type): Use it.
1472         * tree.h (fit_double_type): Export.
1473
1474 2007-01-08  Jan Hubicka  <jh@suse.cz>
1475
1476         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
1477         * ipa.c (function_and_variable_visibility): Fix return type.
1478
1479 2007-01-08  Richard Guenther  <rguenther@suse.de>
1480
1481         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
1482         of offset to build the index.
1483         * tree-pretty-print.c (dump_generic_node): Don't build negated
1484         const just for printing.
1485         * c-pretty-print.c (pp_c_integer_constant): Likewise.
1486         * builtins.c (fold_builtin_int_roundingfn): Check if result
1487         fits the type by using force_fit_type and comparing the result.
1488         * predict.c (predict_loops): Use compare_tree_int for comparison.
1489         * tree.c (build_int_cst): Fall back to integer_type_node for
1490         NULL_TREE type.
1491         (build_int_cst_wide): Assert type is non-null.
1492
1493 2007-01-08  Roberto Costa  <roberto.costa@st.com>
1494
1495         * tree-vrp.c (extract_range_from_cond_expr): New.
1496         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
1497         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
1498         as expressions.
1499         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
1500         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
1501         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
1502         used as expressions.
1503         * tree-object-size.c (cond_expr_object_size): New.
1504         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
1505
1506 2007-01-08  Jan Hubicka  <jh@suse.cz>
1507
1508         * tree-ssa-forwprop.c (forward_propagate_into_cond,
1509         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
1510         statement.
1511
1512 2007-01-08  Richard Guenther  <rguenther@suse.de>
1513
1514         PR tree-optimization/23603
1515         * tree-vrp.c (set_value_range_to_truthvalue): New function.
1516         (extract_range_from_binary): Fall back to truthvalue instead of
1517         varying for TRUTH_*_EXPR.
1518         (extract_range_from_comparison): Fall back to truthvalue instead of
1519         varying.
1520         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
1521         if all visited PHI values were constant.
1522
1523 2007-01-08  Jan Hubicka  <jh@suse.cz>
1524
1525         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
1526         (cgraph_analyze_function): break out from ...
1527         (cgraph_finalize_compilation_unit): ... here.
1528         (cgraph_expand_function): Remove forgoten commented out line.
1529         (cgraph_optimize): Analyze functions.
1530
1531 2007-01-08  Jan Hubicka  <jh@suse.cz>
1532
1533         * tree-pas.h (TODO_remove_function): New flag.
1534         (TODO_update*): Renumber.
1535         (pass_ipa_increase_alignment,
1536         pass_ipa_function_and_variable_visibility): New passes.
1537         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
1538         (cgraph_function_and_variable_visibility): Move to ipa.c
1539         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
1540         cgraph_increase_alignment.
1541         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
1542         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
1543         calling tree-inline.
1544         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
1545         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
1546         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
1547         (gate_increase_alignment): New function.
1548         (pass_ipa_increase_alignment): New pass.
1549         * ipa.c: Inline tree-pass.h and timevar.h
1550         (function_and_variable_visibility): Move here from cgraphunit.c
1551         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
1552         * passes.c (init_optimization_passes): Add the two new passes.
1553         (execute_todo): Handle cgraph_remove_functions.
1554
1555 2007-01-08  Nick Clifton  <nickc@redhat.com>
1556
1557         * config/frv/predicates.md (reg_or_0_operand): Accept
1558         CONST_DOUBLEs.
1559
1560 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
1561
1562         * config/bfin/rtems.h, config/bfin/t-rtems: New.
1563         * config.gcc: Add bfin*-rtems*.
1564
1565 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
1566
1567         * c.opt: Add -flax-vector-conversions.
1568         * c-typeck.c (convert_for_assignment): Pass flag to
1569         vector_types_convertible_p to allow emission of note.
1570         (digest_init): Likewise.
1571         * c-opts.c: Handle -flax-vector-conversions.
1572         * c-common.c (flag_lax_vector_conversions): New.
1573         (vector_types_convertible_p): Unless -flax-vector conversions
1574         has been passed, disallow conversions between vectors with
1575         differing numbers of subparts and/or element types.  If such
1576         a conversion is disallowed, possibly emit a note on the first
1577         occasion only to inform the user of -flax-vector-conversions.
1578         The new last argument specifies this.
1579         * c-common.h (flag_lax_vector_conversions): New.
1580         (vector_types_convertible_p): Add extra argument.
1581         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
1582         char_type_node for V*QI type vectors.
1583         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
1584         Update to satisfy new typechecking rules.
1585         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
1586         C and C++ variants.
1587         * doc/invoke.texi (C Dialect Options): Document
1588         -flax-vector-conversions.
1589
1590 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
1591
1592         PR tree-optimization/29877
1593         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
1594         a register variable on the RHS to not be replaceable.
1595
1596 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
1597         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
1598         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
1599         const_simm15, const_pow2, const_npow2): Added.
1600         * config/score/misc.md (insv, extv, extzv, movmemsi,
1601         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
1602         * config/score/score.c (score_address_cost, score_select_cc_mode):
1603         Added.
1604         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
1605         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
1606         Update score_print_operand.
1607         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
1608         Adjust register allocate order and update some macro define.
1609         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
1610         mdx_block_move_straight, mdx_block_move_loop_head,
1611         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
1612         mdx_block_move): Added.
1613         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
1614         mdp_limm): Updated and fix some bug and typo.
1615         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
1616         (movsf, movdf, doloop_end): Added.
1617
1618 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
1619
1620         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
1621         config/arm/thumb2.md: Fix comment typos.
1622         * doc/extend.texi: Fix a typo.
1623
1624 2007-01-07  Eric Christopher  <echristo@apple.com>
1625
1626         * configure.ac: Check for __stack_chk_fail for darwin.
1627         * configure: Regenerate.
1628
1629 2007-01-07  Richard Guenther  <rguenther@suse.de>
1630
1631         * tree-vrp.c (extract_range_from_assert): CSE calls to
1632         compare_values where possible.
1633         (extract_range_from_unary_expr): Likewise.
1634
1635 2007-01-07  Anatoly Sokolov <aesok@post.ru>
1636
1637         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
1638         * config/avr/avr.c (call_insn_operand): Delete function.
1639         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
1640         instead of register number. Use predicates.
1641         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
1642         call_insn_operand): Add.
1643
1644 2007-01-06  Jan Hubicka  <jh@suse.cz>
1645
1646         * tree-pass.h (pass_build_cgraph_edges): Declare.
1647         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
1648         (visited_nodes): Remove.
1649         (cgraph_create_edges): Move to cgraphbuild.c; rename to
1650         build_cgrpah_edges; make visited_nodes local.
1651         (cgraph_process_new_functions): DO not call initialize_inline_failed.
1652         (record_references_in_initializer): Move to cgraphbuild.c
1653         (initialize_inline_failed, rebuild_cgraph_edges,
1654         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
1655         (verify_cgraph_node): Make visited_nodes local.
1656         (cgraph_analyze_function): Do not call cgraph_create_edges and
1657         initialize_inline_failed.
1658         (cgraph_expand_function): Do not call cgraph_lower_function;
1659         assert that function is already lowered.
1660         * Makefile.in (cgraphbuild.o): New.
1661         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
1662         at the end of lowering passes.
1663
1664 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
1665
1666         * ifcvt.c (cond_move_convert_if_block): New function, code
1667         factored out from...
1668         (cond_move_process_if_block): ...here.  Call the new function
1669         on the THEN and ELSE blocks.
1670         (merge_if_block): Do not copy global_live_at_end, merge_blocks
1671         already takes care of this.
1672
1673 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1674
1675         PR c/19978
1676         * tree.h (TREE_OVERFLOW_P): New.
1677         * c-typeck.c (parser_build_unary_op): Warn only if result
1678         overflowed and operands did not.
1679         (parser_build_binary_op): Likewise.
1680         (convert_for_assignment): Remove redundant overflow_warning.
1681         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
1682
1683 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1684
1685         * c-typeck.c (store_init_value): Split over two lines to follow
1686         the GNU coding style.
1687
1688 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
1689
1690         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
1691         __GXX_EXPERIMENTAL_CXX0X__.
1692         * doc/cpp.texi: Same.
1693
1694 2007-01-05  Richard Guenther  <rguenther@suse.de>
1695
1696         PR middle-end/27826
1697         * tree.c (get_narrower): Do not construct COMPONENT_REFs
1698         with mismatched types.  Instead explicitly build a
1699         conversion NOP_EXPR.
1700
1701 2007-01-05  Ian Lance Taylor  <iant@google.com>
1702
1703         * c-common.c (decl_with_nonnull_addr_p): New function.
1704         (c_common_truthvalue_conversion): Call it.
1705         * c-typeck.c (build_binary_op): Likewise.
1706         * c-common.h (decl_with_nonnull_addr_p): Declare.
1707
1708 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
1709
1710         PR c/30360
1711         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
1712         denom against 0.0.
1713
1714 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
1715
1716         * doc/install.texi (Final install): Document the fact that
1717         the GNAT runtime should not be stripped.
1718
1719 2007-01-04  Jan Hubicka  <jh@suse.cz>
1720
1721         * tree-inline.c (fold_marked_statements): Update operand caches
1722         and EH after folding
1723
1724 2007-01-04  Ian Lance Taylor  <iant@google.com>
1725
1726         * c-common.c (check_function_nonnull): Whitespace fix.
1727
1728 2007-01-04  Jan Hubicka  <jh@suse.cz>
1729
1730         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
1731         comitted older version of patch.
1732         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
1733         function, celanup cfg and collect garbage.
1734
1735 2007-01-04  Mike Stump  <mrs@apple.com>
1736
1737         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
1738         (clean): Likewise.
1739
1740 2007-01-04  Eric Christopher  <echristo@apple.com>
1741
1742         * libgcc2.c (__bswapsi2): Use SItype.
1743         (__bswapdi2): Use DItype.
1744         * libgcc2.h: Update for above.
1745
1746 2007-01-04  Paul Brook  <paul@codesourcery.com>
1747
1748         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
1749         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
1750         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
1751         umulsidi3, umulsidi3adddi, smulsi3_highpart,
1752         umulsi3_highpart): Make conditional on !arm_arch6.
1753         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
1754         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
1755         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
1756         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
1757         umulsi3_highpart_v6): New insns.
1758
1759 2007-01-04  Roger Sayle  <roger@eyesopen.com>
1760
1761         * fold-const.c (fold_convert): When casting an expression to void,
1762         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
1763         have a type.  Instead of attempting to build a NOP_EXPR, return
1764         these "special" trees directly.
1765
1766 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
1767
1768         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
1769         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
1770         rs6000_cost->fp not 0 as adjustment for outer NEG.
1771
1772 2007-01-04  Jan Hubicka  <jh@suse.cz>
1773
1774         * cgraph.c (cgraph_release_function_body): New function.
1775         (cgraph_remove_node): Use it.
1776         * cgraph.h (cgraph_release_function_body): Declare.
1777         * cgraphunit.c (cgraph_expand_function): Use it.
1778         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
1779         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
1780         is initialized and while compilation of other function is running.
1781         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
1782         statement CFG annotations from here to ...
1783         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
1784
1785 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
1786
1787         * cfgloop.h (enum li_flags): Make the constants powers of two.
1788
1789 2007-01-04  Jan Hubicka  <jh@suse.cz>
1790
1791         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
1792         set.
1793         (fold_marked_statements): New function.
1794         (optimize_inline_calls, tree_function_versioning): Fold new statements.
1795         * tree-inline.h (copy_body_data): Add statements_to_fold.
1796
1797 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
1798
1799         * config.gcc: Mention libgcc/config.host.
1800         * Makefile.in: Update comments mentioning libgcc.
1801         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
1802         (all.cross, start.encap, rest.encap, rest.cross): Update
1803         dependencies for libgcc move.
1804         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
1805         (clean-target, clean-target-libgcc): Delete.
1806         (srcdirify, GCC_EXTRA_PARTS): New macros.
1807         (libgcc-support, libgcc.mvars): New rules.
1808         (distclean): Remove mention of mklibgcc.
1809         (install): Don't reference INSTALL_LIBGCC.
1810         (install-common): Don't reference EXTRA_PARTS.
1811         (install-libgcc, install-multilib): Delete rules.
1812         * mklibgcc.in: Delete file.
1813         * doc/configfiles.texi: Don't mention mklibgcc.
1814
1815         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
1816         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
1817         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
1818         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
1819         (mklibgcc, ldblspecs): Likewise.
1820
1821         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
1822         (posixpre.def): Use $(T).
1823         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
1824         (SHLIB_SRC, SHLIB_INSTALL): Delete.
1825         (SHLIB_LINK): Make dummy.
1826         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
1827
1828         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
1829
1830         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
1831         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
1832         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
1833         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
1834
1835         * configure.ac (all_outputs): Remove mklibgcc.
1836         * configure: Regenerated.
1837
1838 2007-01-03  Josh Conner  <jconner@apple.com>
1839
1840         PR middle-end/29683
1841         * calls.c (compute_argument_addresses): Set stack and stack_slot
1842         for partial args, too.
1843         (store_one_arg): Use locate.size.constant for the size when
1844         generating a save_area.
1845
1846 2007-01-03  Robert Kennedy <jimbob@google.com>
1847
1848         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
1849         whose definitions are deleted due to basic block merging.
1850
1851 2007-01-03  Paul Brook  <paul@codesourcery.com>
1852
1853         PR target/16634
1854         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
1855         functions.
1856         (use_return_insn): Return 0 for Thumb interrupt functions.
1857         (print_multi_reg): Add rfe argument for IRQ returns.
1858         (arm_output_epilogue): Pop interrupt return address directly into PC.
1859         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
1860
1861 2007-01-03  Paul Brook  <paul@codesourcery.com>
1862
1863         Merge from sourcerygxx-4_1.
1864         * config/arm/thumb2.md: New file.
1865         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
1866         Thumb-2.
1867         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
1868         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
1869         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
1870         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
1871         tables.
1872         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
1873         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
1874         comptibility.
1875         * config/arm/ieee754-sf.S: Ditto.
1876         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
1877         (thumb1_base_register_rtx_p): ... to this.
1878         (thumb_index_register_rtx_p): Rename...
1879         (thumb1_index_register_rtx_p): ... to this.
1880         (thumb_output_function_prologue): Rename...
1881         (thumb1_output_function_prologue): ... to this.
1882         (thumb_legitimate_address_p): Rename...
1883         (thumb1_legitimate_address_p): ... to this.
1884         (thumb_rtx_costs): Rename...
1885         (thumb1_rtx_costs): ... to this.
1886         (thumb_compute_save_reg_mask): Rename...
1887         (thumb1_compute_save_reg_mask): ... to this.
1888         (thumb_final_prescan_insn): Rename...
1889         (thumb1_final_prescan_insn): ... to this.
1890         (thumb_expand_epilogue): Rename...
1891         (thumb1_expand_epilogue): ... to this.
1892         (arm_unwind_emit_stm): Rename...
1893         (arm_unwind_emit_sequence): ... to this.
1894         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
1895         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
1896         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
1897         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
1898         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
1899         thumb2_output_casesi): New functions.
1900         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
1901         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
1902         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
1903         THUMB2_WORK_REGS): Define.
1904         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
1905         arm_condexec_mask, arm_condexec_masklen)): New variables.
1906         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
1907         (arm_override_options): Check new CPU capabilities.
1908         Set new architecture flag variables.
1909         (arm_isr_value): Handle v7m interrupt functions.
1910         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
1911         Thumb-2.
1912         (const_ok_for_arm): Handle Thumb-2 constants.
1913         (arm_gen_constant): Ditto.  Use movw when available.
1914         (arm_function_ok_for_sibcall): Return false for v7m interrupt
1915         functions.
1916         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
1917         (thumb_find_work_register, arm_load_pic_register,
1918         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
1919         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
1920         print_multi_reg, output_mov_long_double_fpa_from_arm,
1921         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
1922         output_mov_double_fpa_from_arm, output_move_double,
1923         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
1924         output_return_instruction, arm_output_function_prologue,
1925         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
1926         arm_output_mi_thunk, thumb_set_return_address): Ditto.
1927         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
1928         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
1929         (arithmetic_instr, shift_op): Use arm_shift_nmem.
1930         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
1931         '!' and 'L'.
1932         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
1933         (thumb_expand_prologue): Use thumb_set_frame_pointer.
1934         (arm_file_start): Output directive for unified syntax.
1935         (arm_unwind_emit_set): Handle stack alignment instruction.
1936         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
1937         Add v6t2, v7, v7a, v7r and v7m.
1938         (RETLDM): Add Thumb-2 code.
1939         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
1940         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
1941         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
1942         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
1943         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
1944         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
1945         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
1946         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
1947         ADJUST_INSN_LENGTH): Define.
1948         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
1949         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
1950         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
1951         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
1952         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
1953         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
1954         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
1955         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
1956         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
1957         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
1958         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
1959         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
1960         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
1961         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
1962         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
1963         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
1964         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
1965         cortex-m3.
1966         * config/arm/arm-tune.md: Regenerate.
1967         * config/arm/arm-protos.h: Update prototypes.
1968         * config/arm/vfp.md: Enable patterns for Thumb-2.
1969         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
1970         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
1971         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
1972         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
1973         * config/arm/libunwind.S: Add Thumb-2 code.
1974         * config/arm/constraints.md: Update include Thumb-2.
1975         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
1976         * config/arm/ieee754-df.S: Ditto.
1977         * config/arm/bpabi.S: Ditto.
1978         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
1979         * config/arm/predicates.md (low_register_operand,
1980         low_reg_or_int_operand, thumb_16bit_operator): New.
1981         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
1982         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
1983         * config/arm/t-arm-elf: Add armv7 multilib.
1984         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
1985         Include thumb2.md.
1986         (UNSPEC_STACK_ALIGN, ce_count): New.
1987         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
1988         arm_zero_extendsidi2, arm_zero_extendqidi2): New
1989         insns/expanders.
1990         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
1991         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
1992         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
1993         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
1994         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
1995         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
1996         insns.
1997         * doc/extend.texi: Document ARMv7-M interrupt functions.
1998         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
1999
2000 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
2001
2002         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2003         (struct _Unwind_Context): Rename args_size to flags, remove
2004         signal_frame field, add a new args_size field and version field.
2005         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2006         _Unwind_IsExtendedContext): New inline functions.
2007         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2008         Assume by_value array is only present if _Unwind_IsExtendedContext.
2009         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2010         _Unwind_IsSignalFrame.
2011         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2012         (uw_update_context_1): Use _Unwind_SetSignalFrame.
2013         (uw_init_context_1): Initialize context->flags to
2014         EXTENDED_CONTEXT_BIT.
2015         * config/rs6000/linux-unwind.h (frob_update_context): Use
2016         _Unwind_SetSignalFrame.
2017
2018 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
2019
2020         PR middle-end/30353
2021         * gimplify.c (gimplify_modify_expr_complex_part): Move below
2022         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
2023         the value.
2024
2025 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
2026
2027         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2028         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2029         RET, AT_SP): Remove.
2030         * config/i386/i386.md (*sse_prologue_save_insn): Use return
2031         instead of RET.
2032
2033         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2034         * cfgbuild.c (find_basic_blocks): Likewise.
2035         * cfgrtl.c (rtl_create_basic_block): Likewise.
2036         * function.c (temp_slots_at_level): Likewise.
2037         * reg-stack.c (stack_regs_mentioned): Likewise.
2038         * regclass.c (allocate_reg_info): Likewise.
2039         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2040         set_bb_for_stmt, move_block_to_fn): Likewise.
2041         * tree-complex.c (tree_lower_complex): Likewise.
2042         * vec.h (VEC_safe_grow_cleared): New.
2043
2044         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2045
2046 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
2047
2048         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2049         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2050         to loop_version.
2051         * cfgloopmanip.c (scale_loop_frequencies): Export.
2052         (loopify): Scale the frequencies by prescribed coefficients.
2053         (set_zero_probability): New function.
2054         (duplicate_loop_to_header_edge): Improve updating of frequencies.
2055         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2056         and frequencies according to arguments.
2057         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2058         correctly.
2059         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2060         * modulo-sched.c (sms_schedule): Set probabilities for entering
2061         versioned loop correctly.
2062         * tree-vect-transform.c (vect_transform_loop): Ditto.
2063         * cfgloop.h (loopify, loop_version): Declaration changed.
2064         (scale_loop_frequencies): Declared.
2065
2066 2007-01-02  Jan Hubicka  <jh@suse.cz>
2067
2068         * cgraph.c: Include tree-flow.h
2069         (cgraph_add_new-function): Handle IPA_SSA mode; execute
2070         early_local_passes.
2071         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2072         * tree-pass.h (pass_all_early_optimizations): Declare.
2073         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2074         early_local_passes.
2075         (cgraph_analyze_function): Do early_local_passes.
2076         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2077         Do not add referenced vars.
2078         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2079         SSA form.
2080         (gate_all_early_local_passes): New gate.
2081         (pass_early_local_passes): Use new gate.
2082         (execute_early_local_optimizations): New functions.
2083         (gate_all_early_optimizations): New gate.
2084         (pass_all_early_optimizations): New pass.
2085         (execute_free_datastructures): Free SSA only when initialized.
2086         (gate_init_datastructures): Init only when optimizing.
2087         (tree_lowering_passes): Do early local passes when called late.
2088         * tree-profile.c (do_tree_profiling): Don't profile functions added
2089         late.
2090         (do_early_tree_profiling, pass_early_tree_profile): Kill.
2091         * tree-cfg.c (update_modified_stmts): Do not update when operands are
2092         not active.
2093         * passes.c (init_optimizations_passes): Reorder so we go into SSA
2094         during early_local_passes.
2095         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
2096
2097
2098 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
2099
2100         * Makefile.in: Update copyright year.
2101
2102 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
2103
2104         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
2105         in $(lang_checks) and check-consistency targets.
2106
2107 2007-01-02  Jan Hubicka  <jh@suse.cz>
2108
2109         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2110         Do not add referenced vars.
2111         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
2112         are not active.
2113         * passes.c (init_optimization_passes): Put mudflap_2 after
2114         free_datastructures.
2115
2116 2007-01-02  Jan Hubicka  <jh@suse.cz>
2117
2118         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
2119         Set NOTHROW flag on call statements proved to be nothrow.
2120         Update statement of local calls so new pure/const functions are
2121         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
2122         goto receivers.
2123         (tree_rest_of_compilation): Register hooks and initialize bitmap
2124         early. Do not set after_inlining flag.
2125
2126 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
2127
2128         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
2129         HOST_BITS_PER_LONGLONG
2130
2131 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2132
2133         PR c/19977
2134         * c-typeck.c (store_init_value): Don't emit pedantic overflow
2135         warning for non-static initializers.
2136
2137 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
2138
2139         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
2140         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
2141         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
2142         Always use set_unique_reg_note to add REG_EQUAL notes.
2143
2144 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
2145
2146         Revert:
2147         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
2148
2149         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2150         * cfgbuild.c (find_basic_blocks): Likewise.
2151         * cfgrtl.c (rtl_create_basic_block): Likewise.
2152         * function.c (temp_slots_at_level): Likewise.
2153         * reg-stack.c (stack_regs_mentioned): Likewise.
2154         * regclass.c (allocate_reg_info): Likewise.
2155         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2156         set_bb_for_stmt, move_block_to_fn): Likewise.
2157         * tree-complex.c (tree_lower_complex): Likewise.
2158         * vec.h (VEC_safe_grow_cleared): New.
2159
2160 2007-01-02  Ian Lance Taylor  <iant@google.com>
2161
2162         * c-common.c (c_common_truthvalue_conversion): When warning about
2163         using an assignment as a truth value, set TREE_NO_WARNING.
2164
2165 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2166
2167         PR middle-end/7651
2168         * c.opt (Wold-style-declaration): New.
2169         * doc/invoke.texi (C-only Warning Options): New.
2170         (Wold-style-declaration): Document it.
2171         (Wextra): Enabled by -Wextra.
2172         * c-opts.c (c_common_post_options): Enabled by -Wextra.
2173         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
2174         -Wold-style-declaration.
2175
2176 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
2177
2178         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2179         * cfgbuild.c (find_basic_blocks): Likewise.
2180         * cfgrtl.c (rtl_create_basic_block): Likewise.
2181         * function.c (temp_slots_at_level): Likewise.
2182         * reg-stack.c (stack_regs_mentioned): Likewise.
2183         * regclass.c (allocate_reg_info): Likewise.
2184         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2185         set_bb_for_stmt, move_block_to_fn): Likewise.
2186         * tree-complex.c (tree_lower_complex): Likewise.
2187         * vec.h (VEC_safe_grow_cleared): New.
2188
2189 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
2190
2191         * c-common.c(c_common_nodes_and_builtins): Since variants of
2192         void_type_node get built before it is given a name, we need to
2193         give those variants the name, too.
2194         (complete_array_type): We need to work with the canonical main
2195         type of the array, from which we will build the qualified version.
2196         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
2197         * print-tree.c (print_node): Display canonical type information
2198         for each type.
2199         * stor-layout.c (layout_type): When we don't know the
2200         alignment of a type for which we're building an array, we end up
2201         guessing wrong, so make the type require structural equality.
2202         * tree.c (make_node_stat): When we build a new type, it is its
2203         own canonical type.
2204         (build_type_attribute_qual_variant): When building an attribute
2205         variant, its canonical type is the non-attribute variant. However,
2206         if the attributes are target-dependent and they differ, we need to
2207         use structural equality checks for this type.
2208         (build_qualified_type): A qualified type is not equivalent to its
2209         unqualified variant; set the canonical type appropriately.
2210         (build_distinct_type_copy): When building a distinct type from
2211         another type, the new type is its own canonical type.
2212         (build_variant_type_copy): When building a new type variant, we
2213         assume that it is equivalent to the original type.
2214         (build_pointer_type_for_mode): When building a pointer type, also
2215         build a canonical type pointer.
2216         (build_reference_type_for_mode): When building a reference type,
2217         also build a canonical type reference.
2218         (build_index_type): When we can't hash an index type (e.g.,
2219         because its maximum value is negative), the index type requires
2220         structural equality tests.
2221         (build_array_type): Build the canonical form of an array type.
2222         (build_function_type): Function types require structural equality,
2223         because they contain default arguments, attributes, etc.
2224         (build_method_type_directly): Ditto for method types.
2225         (build_offset_type): Build the canonical offset type.
2226         (build_complex_type): Build the canonical vector type.
2227         (make_vector_type): Build the canonical vector type.
2228         * tree.h (TYPE_CANONICAL): New.
2229         (TYPE_STRUCTURAL_EQUALITY_P): New.
2230         (SET_TYPE_STRUCTURAL_EQUALITY): New.
2231         (struct tree_type): Added "canonical" field.
2232         * params.h (VERIFY_CANONICAL_TYPES): New.
2233         * doc/c-tree.texi (TYPE_CANONICAL): Document.
2234         (TYPE_STRUCTURAL_EQUALITY_P): Document.
2235         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
2236         * doc/invoke.texi (verify-canonical-types): Document --param
2237         parameter for verifying canonical types.
2238
2239 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
2240
2241         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
2242         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
2243         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
2244         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
2245         tm_file.
2246         * config/rs6000/e500.h: New.
2247         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
2248         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
2249         Remove.
2250         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
2251         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
2252         TARGET_E500_DOUBLE): Remove.
2253         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
2254         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
2255         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
2256         * config/rs6000/rs6000.c (rs6000_override_options): Use
2257         CHECK_E500_OPTIONS.
2258
2259 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
2260
2261         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
2262         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
2263         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
2264         rs6000_emit_cmove): Don't check TARGET_E500.
2265         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
2266         sordered): Don't check TARGET_E500.
2267
2268 2007-01-01  Eric Christopher  <echristo@apple.com>
2269
2270         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
2271         HARD_REGISTER_NUM_P.
2272
2273 2007-01-01  Roger Sayle  <roger@eyesopen.com>
2274
2275         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
2276         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
2277         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
2278         eq/ne Y".
2279
2280 2007-01-01  Mike Stump  <mrs@apple.com>
2281
2282         * configure.ac: Remove support for building with Apple's gcc-3.1.
2283
2284 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
2285
2286         PR middle-end/30311
2287         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
2288         REG is a hard register.  Do not modify REG before calling
2289         subreg_nregs.
2290         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
2291         multiple of the size of YMODE for certain lowpart cases.
2292
2293 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
2294
2295         PR middle-end/30253
2296         * gimplify (voidify_wrapper_expr): Update for
2297         GIMPLE_MODIFY_STMT.
2298
2299 2007-01-01  Andreas Schwab  <schwab@suse.de>
2300
2301         PR target/29166
2302         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
2303         save of BR0 in extra_spill_size instead of spill_size.
2304         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
2305         area.
2306         (ia64_expand_epilogue): Restore BR0 from its new location.
2307
2308 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
2309
2310         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
2311         Use a temporary variable if the left hand side is not a gimple
2312         register.
2313
2314 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
2315
2316         * gimplify.c (gimplify_return_expr): Make the temporary variable
2317         for the return expression, a gimple register variable.
2318
2319 2007-01-01  Jan Hubicka  <jh@suse.cz>
2320
2321         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
2322         INSN_LIST.
2323
2324 2007-01-01  Mike Stump  <mrs@apple.com>
2325
2326         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
2327         .literal16.
2328         * config/darwin.c (machopic_select_rtx_section): Use
2329         HAVE_GAS_LITERAL16.
2330         (darwin_mergeable_constant_section): Likewise.
2331         * configure: Regenerate.
2332         * config.in: Regenerate.
2333
2334 2007-01-01  Jan Hubicka  <jh@suse.cz>
2335             Andrew Pinski  <pinskia@gmail.com>
2336
2337         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
2338         before starting IPA passes.