OSDN Git Service

* fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-01-29  Roger Sayle  <roger@eyesopen.com>
2             Richard Guenther  <rguenther@suse.de>
3
4         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
5         avoid undefined behaviour on overflow.  Use force_fit_type_double
6         to construct the constant with the specified TREE_OVERFLOW.
7
8 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
9
10         * config/dfp-bit.c: Add parameterized support for fp exceptions.
11         * config/dfp-bit.h: Ditto.
12
13 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
14
15         * c-decl.c (pop_scope): Replace warnings with call to
16         warn_for_unused_label.
17         * c-common.h (warn_for_unused_label): Declare.
18         * c-common.c (warn_for_unused_label): Define.
19
20 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
21
22         * tree-optimize.c (update_inlined_to_pointers): Delete unused
23         function.
24
25 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
26
27         * Makefile.in (USER_H): Remove decfloat.h.
28         * ginclude/decfloat.h: Delete, moving contents to ...
29         * ginclude/float.h: Add support for decimal floating point,
30         guarded by __STDC_WANT_DEC_FP__.
31
32 2007-01-29  Mike Stump  <mrs@apple.com>
33
34         * doc/gccint.texi (Top): Rename Loop Representation to Loop
35         Analysis and Representation to resolve case insensitive conflict.
36         * doc/loop.texi (Loop Analysis and Representation): Likewise.
37
38 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
39
40         * tree.h (struct tree_memory_tag): Add aliases member.
41         (MTAG_ALIASES): New macro.
42         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
43         (add_may_alias): Remove pointer-set. Update for may_aliases being
44         a bitmap. 
45         (mark_aliases_call_clobbered): Update for may_aliases being a
46         bitmap.
47         (compute_tag_properties): Ditto.
48         (create_partition_for): Ditto.
49         (compute_memory_partitions): Ditto.
50         (dump_may_aliases_for): Ditto.
51         (is_aliased_with): Ditto.
52         (add_may_alias_for_new_tag): Ditto.
53         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
54         (compute_is_aliased): New function.
55         (compute_may_aliases): Call compute_is_aliased).
56         (init_alias_info): Initialize alias_bitmap_obstack.
57         (union_alias_set_into): New function.
58         (compute_flow_sensitive_aliasing): Use union_aliases_into.
59         (have_common_aliases_p): Rewrite to take two bitmaps and use
60         intersection.
61         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
62         Update for bitmaps.
63         (finalize_ref_all_pointers): Update for add_may_alias changes.
64         (new_type_alias): Ditto.
65         * tree-flow-inline.h (may_aliases): Return a bitmap.
66         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
67         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
68         may_aliases being a bitmap.
69         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
70         may_aliases now returns a bitmap.
71         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
72         being a bitmap.
73         * tree-ssa-operands.c (add_virtual_operand): Update for
74         may_aliases being a bitmap.
75         
76 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
77
78         PR tree-optimization/30630
79         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
80         variable as changed.
81
82 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
83
84         PR c++/28266
85         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
86         expanded only once even if an error occurs.
87
88 2007-01-29  Ben Elliston  <bje@au.ibm.com>
89
90         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
91         (__gcov_average_profiler): Likewise.
92         (__gcov_ior_profiler): Likewise.
93         (__gcov_merge_ior): Likewise.
94
95 2007-01-28  Jan Hubicka  <jh@suse.cz>
96
97         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
98         * value-prof.c (stringop_block_profile): Likewise.
99
100 2007-01-28  Jan Hubicka  <jh@suse.cz>
101
102         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
103         variant handling histograms; add wrapper.
104         (clear_storage_via_libcall): Export.
105         (emit_block_move_hints): Break out from ...; add histograms.
106         (emit_block_move): ... this one.
107         (clear_storage_hints): Break out from ...; add histograms.
108         (clear_storage): ... this one.
109         (set_storage_via_memset): Handle histogram.
110         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
111         (clear_storage_hints, clear_storage_via_libcall): Declare.
112         (set_storage_via_setmem): Update prototype.
113         * doc/md.texi (movmem, setmem): Document new arguments.
114
115         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
116         new histograms.
117         (stringop_block_profile): New global function.
118         (tree_stringops_values_to_profile): Profile block size and alignment.
119         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
120         HIST_TYPE_IOR.
121         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
122         (stringop_block_profile): Declare.
123         * builtins.c: Include value-prof.h.
124         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
125         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
126         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
127         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
128         * profile.c (instrument_values): Add new counters.
129         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
130         calls.
131         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
132         (tree_init_edge_profiler): Build new profilers.
133         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
134         (pass_tree_profile): Add dump.
135         (tree_profile_hooks): Update.
136         * Makefile.in (LIBGCOV): Add new constants.
137         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
138         __gcov_ior_profiler): New.
139         * i386.md (movmem/setmem expanders): Add new optional arguments.
140
141 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
142
143         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
144
145 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
146
147         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
148         (tree_unswitch_single_loop): Pass eni_size_weights to
149         tree_num_loop_insns.
150         * tree-ssa-loop-manip.c: Include tree-inline.h.
151         (can_unroll_loop_p): Pass eni_size_weights to
152         tree_num_loop_insns.
153         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
154         Pass eni_size_weights to estimate_num_insns.
155         * tree.h (init_inline_once): Export.
156         * toplev.c (backend_init): Call init_inline_once.
157         * cgraphunit.c (cgraph_process_new_functions,
158         cgraph_analyze_function): Pass eni_inlining_weights to
159         estimate_num_insns.
160         * ipa-inline.c (compute_inline_parameters): Ditto.
161         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
162         to estimate_num_insns.
163         (try_unroll_loop_completely): Pass eni_size_weights to
164         tree_num_loop_insns.
165         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
166         ot estimate_num_insns.
167         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
168         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
169         * tree-inline.c (eni_inlining_weights, eni_size_weights,
170         eni_time_weights): New variables.
171         (init_inline_once): Initialize them.
172         (struct eni_data): Mew.
173         (estimate_num_insns_1, estimate_num_insns): Use weights.
174         * tree-inline.h (struct eni_weights_d): New.
175         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
176         (estimate_num_insns): Declaration changed.
177         * cfgloop.h (tree_num_loop_insns): Declaration changed.
178         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
179         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
180
181 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
182
183         * tree-data-ref.c (conflict_fn): Assert that the number of affine
184         relations in the conflict function is valid.
185
186 2007-01-27  Ian Lance Taylor  <iant@google.com>
187
188         * common.opt: Add fstrict-overflow.
189         * opts.c (decode_options): Set flag_strict_overflow if -O2.
190         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
191         (TYPE_OVERFLOW_UNDEFINED): Define.
192         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
193         Replace all uses.
194         * tree.h (TYPE_TRAP_SIGNED): Don't define.
195         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
196         (fold_negate_expr): Likewise.
197         (make_range): Likewise.
198         (extract_muldiv_1): Likewise.
199         (maybe_canonicalize_comparison): Likewise.
200         (fold_comparison): Likewise.
201         (fold_binary): Likewise.
202         (tree_expr_nonnegative_p): Likewise.
203         (tree_expr_nonzero_p): Likewise.
204         * tree-vrp.c (compare_values): Likewise.
205         (extract_range_from_binary_expr): Likewise.
206         (extract_range_from_unary_expr): Likewise.
207         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
208         Likewise.
209         (nowrap_type_p): Likewise.
210         * tree-scalar-evolution.c (simple_iv): Likewise.
211         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
212         (build_range_check): Likewise.
213         (extract_muldiv_1): Likewise.
214         (fold_comparison): Likewise.
215         * tree-vrp.c (vrp_int_const_binop): Likewise.
216         (extract_range_from_unary_expr): Likewise.
217         * convert.c (convert_to_integer): Likewise.
218         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
219         (fold_comparison): Likewise.
220         (fold_binary): Likewise.
221         * optabs.c (optab_for_tree_code): Likewise.
222         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
223         * simplify-rtx.c (simplify_const_relational_operation): Check
224         flag_strict_overflow and flag_trapv.
225         (simplify_const_relational_operation): Likewise.
226         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
227         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
228         -fstrict-overflow.
229
230 2007-01-27  Roger Sayle  <roger@eyesopen.com>
231
232         * tree.c (tree_fold_gcd): Delete.
233         * tree.h (tree_fold_gcd): Remove prototype.
234         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
235         test whether one constant integer is a multiple of another.  Instead
236         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
237         * fold-const.c (multiple_of_p):  We've determined both TOP and
238         BOTTOM are integer constants so we can call int_const_binop directly
239         instead of the more generic const_binop.
240
241 2007-01-27  Roger Sayle  <roger@eyesopen.com>
242
243         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
244         1*X check that the constant hasn't overflowed, to preserve the
245         TREE_OVERFLOW bit.
246         (round_up): Provide an efficient implementation when rouding-up an
247         INTEGER_CST to a power-of-two.
248
249 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
250
251         * doc/sourcebuild.texi: Add comma for clarity.
252         * doc/extend.texi: Fix some typos.
253         * doc/passes.texi: Likewise.
254         * doc/cppinternals.texi: Likewise.
255         * doc/c-tree.texi: Likewise.
256         * doc/tree-ssa.texi: Likewise.
257         * doc/install.texi: Likewise.
258
259 2007-01-27  Jan Hubicka  <jh@suse.cz>
260
261         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
262         (pass_sra): Do not require alias information.
263         * passes.c (init_optimization_passes): Add SRA
264
265 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
266
267         * tracer.c (rest_of_handle_tracer): We already cleaned
268         up the CFG in tracer() so don't do it here again.
269         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
270         cleanup_cfg here, either.  And don't call renumber_insns.
271
272         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
273         and BB_HEAD are in the insn stream here.  Instead make sure
274         that BB_INSN is valid on all insns.  Also, do check here that
275         there are no pending branch predictions...
276         (rtl_verify_flow_info): ...instead of doing it here.  Checks
277         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
278         here.
279
280 2007-01-26  Roger Sayle  <roger@eyesopen.com>
281
282         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
283         function to simplify/factorize operand order canonicalization.
284         (ix86_fixup_binary_operands): Reorganize using the above function.
285         (ix86_binary_operator_ok): Likewise.
286
287 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
288
289         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
290         file scope from optimize_attrs.
291         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
292         up search.
293         (optimize_attrs): Clear insn_code_values after freeing it.
294
295 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
296
297         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
298         a parts.base assignment.
299
300 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
301
302         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
303         (compute_subscript_distance, initialize_data_dependence_relation,
304         finalize_ddr_dependent, analyze_ziv_subscript,
305         analyze_siv_subscript_cst_affine,
306         compute_overlap_steps_for_affine_univar,
307         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
308         analyze_siv_subscript, analyze_miv_subscript,
309         analyze_overlapping_iterations, subscript_dependence_tester_1,
310         compute_self_dependence, free_dependence_relation): Work
311         with affine_fn instead of chrecs.
312         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
313         common_affine_function, affine_function_base,
314         affine_function_constant_p, affine_fn_op, affine_fn_plus,
315         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
316         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
317         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
318         (all_chrecs_equal_p): Removed.
319         * tree-data-ref.h (affine_fn, conflict_function): New types.
320         (struct subscript): Change type of conflicting_iterations_in_a
321         and conflicting_iterations_in_b.
322
323 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
324
325         PR other/30182
326         * config/pa/pa.h (TARGET_HPUX_11): New.
327         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
328         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
329
330 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
331         Richard Guenther  <rguenther@suse.de>
332
333         * tree-ssa-structalias.c (solve_graph): Handle case
334         we merged the variable to another.
335
336 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
337
338         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
339         functions.
340
341         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
342         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
343
344 2007-01-25  DJ Delorie  <dj@redhat.com>
345
346         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
347         allow changes to modes which don't fit in those registers.
348
349         * reload1.c (choose_reload_regs): Check for invalid subregs before
350         computing their locations, not after.
351
352 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
353
354         PR 25127
355         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
356         save Altivec registers in an eh_return function.
357         (compute_vrsave_mask): Likewise.
358         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
359         for saving Altivec registers.
360         (rs6000_emit_prologue): Don't allocate stack twice in
361         eh_return function.  Correct expected value of altivec_save_offset
362         when using save_world.  Describe save of R0 to stack when using
363         save_world.  Describe stack pointer adjustment when using
364         save_world.  Remove duplicated eh_return parameter register saving.
365         Update sp_offset variable after save_world.
366         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
367         darwin-world.asm.
368         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
369         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
370         implies Altivec.
371
372 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
373
374         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
375
376 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
377
378         PR other/30182
379         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
380         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
381
382 2007-01-25  Richard Guenther  <rguenther@suse.de>
383
384         * doc/invoke.texi (-Wcoverage-mismatch): Document.
385         * common.opt (-Wcoverage-mismatch): New warning option.
386         * coverage.c (get_coverage_counts): Ignore coverage mismatch
387         if -Wcoverage-mismatch is given.
388
389 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
390
391         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA. 
392         (ipcp_driver): Change to static definition. 
393         Add dumping of the ifunctions.
394         (constant_val_insert): Remove unused parameter. Support for SSA.
395         (ipcp_propagate_const): Support for SSA.
396         (ipcp_profile_bb_print): Print only analyzed nodes.
397         (ipcp_replace_map_create): Remove support for Fortran constant 
398         for now.
399         * ipa-prop.c (ipa_method_modify_stmt, 
400         ipa_callsite_compute_param): Support for SSA.
401         * ipa-prop.h (ipcp_driver): Remove declaration.
402         (IS_VALID_TREE_MAP_INDEX): Add define.
403
404 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
405
406         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
407
408 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
409
410         PR middle-end/30447
411         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
412         constant argument to cabs and do it without checking for
413         -funsafe-math-optimizations.
414
415 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
416
417         * c-common.h (RID_FIRST_CXX0X): New.
418         (RID_LAST_CXX0X): New.
419         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
420         by -Wall.
421         * c.opt (Wc++0x-compat): New.
422         * doc/invoke.texi (-Wc++0x-compat): Document.
423
424 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
425
426         * builtins.c (fold_builtin_carg): New.
427         (fold_builtin_1): Use it.
428
429 2007-01-24  Jan Hubicka  <jh@suse.cz>
430
431         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
432
433         * ipa-inline.c (initial_insns, max_insns): Delete.
434         (compute_max_insns): New function.
435         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
436         of insns as base for code growth.
437         (cgraph_decide_inlining): Make initial_insns local; do not compute
438         max_insns.
439         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
440         * doc/invoke.texi (inline-unit-growth): Update docs.
441
442 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
443
444         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
445         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
446         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
447         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
448         for CPUs that have PTA_CX16 set.
449
450         PR middle-end/27416
451         * gimplify.c (omp_check_private): New function.
452         (gimplify_scan_omp_clauses): Use it for
453         firstprivate/lastprivate/reduction.
454
455         PR middle-end/30494
456         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
457         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
458
459         PR middle-end/30421
460         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
461         If lower_lastprivate_clauses emits some statements, append them
462         to dlist rather than body_p and to body_p append an initializer.
463         (lower_omp_for): Adjust caller.
464
465 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
466
467         * target.h (globalize_decl_name): New.
468         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
469         * output.h (default_globalize_decl_name): New.
470         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
471         globalize_label.
472         (globalize_decl): Ditto.
473         (default_globalize_decl_name): New.
474         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
475         (ia64_handle_version_id_attribute): New.
476         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
477         (ia64_asm_output_external): Use globalize_decl_name instead
478         of globalize_label.
479         * doc/extend.texi (version_id): New pragma.
480         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
481
482 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
483
484         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with 
485         _uleb128_t and _Unwind_SWord with _sleb128_t.
486         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
487         uw_frame_state_for, uw_update_context_1): Likewise.
488         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
489         * unwind-pe.h (read_uleb128, read_sleb128,
490         read_encoded_value_with_base): Likewise.
491         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
492
493 2007-01-24  Richard Guenther  <rguenther@suse.de>
494
495         * builtins.c (expand_builtin_cexpi): Get the fndecl
496         for cexp in the correct way.
497
498 2007-01-24  Jan Hubicka  <jh@suse.cz>
499
500         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
501
502 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
503
504         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
505         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
506         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
507         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
508         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
509         __SIZEOF_PTRDIFF_T__.
510         * doc/cpp.texi: Documentation for the new builtin macros added.
511
512 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
513
514         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
515         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
516         instead of operands[2] to avoid access past the end of array.
517
518 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
519
520         * reload1.c (emit_reload_insns): Pass the reload register
521         for a non-spill output reload through forget_old_reloads_1.
522
523 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
524
525         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
526         _SOFT_DOUBLE if doubles use software floating-point.
527         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
528         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
529         * config/rs6000/darwin-ldouble.c: Also compile functions for
530         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
531         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
532         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
533         commented-out long double override.
534         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
535         * config/rs6000/eabispe.h: Likewise.
536         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
537         long double for non-SPE.
538         (rs6000_handle_option): Likewise.
539         (invalid_e500_subreg): Disallow more subregs involding DImode,
540         DFmode, TImode or TFmode.
541         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
542         E500 double.
543         (legitimate_lo_sum_address_p): Also check for TFmode for E500
544         double.
545         (rs6000_legitimize_address): Also handle TFmode for E500 double.
546         (rs6000_legitimize_reload_address): Also handle TFmode for E500
547         double.
548         (rs6000_legitimate_address): Also check for TFmode for E500
549         double.
550         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
551         (spe_build_register_parallel): Handle TFmode and TCmode.
552         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
553         double.
554         (function_arg): Handle TFmode and TCmode for E500 double.
555         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
556         in general.
557         (print_operand): Handle TFmode and TImode for %y.
558         (rs6000_generate_compare): Handle TFmode comparisons for E500
559         double.
560         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
561         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
562         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
563         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
564         TFmode for E500 double.
565         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
566         (floatsidf2): Enable for E500 double.
567         (movtf_softfloat): Use rs6000_nonimmediate_operand.
568         (extenddftf2): Change to extenddftf2_fprs.
569         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
570         depending on TARGET_E500_DOUBLE.
571         (extendsftf2): Enable for E500 double.
572         (trunctfdf2): Enable for E500 double.
573         (trunctfsf2): Change to trunctfsf2_fprs.
574         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
575         depending on TARGET_E500_DOUBLE.
576         (floatsitf2): Enable for E500 double.
577         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
578         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
579         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
580         (negtf2): Change to negtf2_internal.
581         (negtf2): New expander.
582         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
583         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
584         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
585         (movdi_internal32): Use rs6000_nonimmediate_operand.
586         (unnamed splitter): Likewise.
587         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
588         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
589         (SPE64TF, DITI): New mode macros.
590         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
591         modes.
592         (frob_tf_ti): New.
593         (frob_<mode>_di_2): New.
594         (frob_tf_di_8_2): New.
595         (frob_di_df): Change to frob_di_<mode>; allow more modes.
596         (frob_ti_tf): New.
597         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
598         more modes.
599         (frob_ti_<mode>_8_2): New.
600         (frob_ti_tf_2): New.
601         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
602         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
603         TFmode.
604         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
605         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
606         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
607         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
608         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
609         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
610         tsttflt_gp): New.
611
612 2007-01-23  Ian Lance Taylor  <iant@google.com>
613
614         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
615         insn-*.o first.
616         (OBJS-archive): Reformat, alphabetize.
617         (OBJS): Change out_object_file to OBJS-md.
618
619 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
620
621         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
622         JUMP_P and CALL_P predicates where applicable.
623         * config/i386/i386.c: Ditto.
624         * config/i386/i386.md: Ditto.
625         * config/i386/mmx.md: Ditto.
626         * config/i386/predicates.md: Ditto.
627
628 2007-01-22  Andreas Schwab  <schwab@suse.de>
629
630         * config/m68k/m68k.h: Fix comment.
631
632 2007-01-22  Jan Hubicka  <jh@suse.cz>
633
634         * passes.c (init_optimization_passes): Do not rerun
635         pass_early_warn_uninitialized.
636
637 2007-01-22  Richard Guenther  <rguenther@suse.de>
638
639         PR tree-optimization/30038
640         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
641         function.
642         (execute_cse_sincos_1): Likewise.
643         (execute_cse_sincos): Likewise.
644         (gate_cse_sincos): Likewise.
645         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
646         the cexpi() canonicalization of sincos().
647         * tree-pass.h (pass_cse_sincos): Declare.
648         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
649
650 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
651
652         PR rtl-optimization/29329
653         * combine.c (replaced_rhs_insn): Rename to i2mod.
654         (replaced_rhs_value): Rename to i2mod_new_rhs.
655         (i2mod_old_rhs): New global variable.
656         (combine_instructions): Adjust for above change.  Save a copy of
657         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
658         note are substituted in the second instruction.
659         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
660         ditch the note if it pertains to the second eliminated register
661         and this register is mentioned in i2mod_old_rhs.
662
663         Revert:
664         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
665
666         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
667         the insn to which the note was originally attached.
668
669 2007-01-21  Jan Hubicka  <jh@suse.cz>
670
671         * ipa-inline.c (inlining_mode): Comment, move up.
672         (cgraph_decide_inlining_incrementally): Do not perform inlining
673         itself; fix handling of flattening of self recursive functions.
674         (cgraph_find_cycles): Remove.
675         (cgraph_flatten_node): Remove.
676         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
677         (try_inline): New function.
678         (cgraph_early_inlining): Update call of
679         cgraph_decide_inlining_incrementally.  Apply inlining here.
680         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
681
682 2007-01-21  Dirk Mueller  <dmueller@suse.de>
683
684         PR bootstrap/30511
685         * tree-vrp.c (check_array_bounds): do not warn
686         about ADDR_EXPR's of ARRAY_REF's which are immediately
687         used in binary expressions.
688
689 2007-01-21  Ira Rosen  <irar@il.ibm.com>
690
691         * tree-vectorizer.h (struct _stmt_vec_info): Add new field 
692         read_write_dep and macros for its access.
693         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
694         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove 
695         argument, call vect_check_interleaving for every independent pair of
696         data-refs. Mark loads that access the same memory location as a store
697         in the loop.
698         (vect_check_dependences): Remove.
699         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences 
700         call, fix the call to vect_analyze_data_ref_dependence.
701         (vect_analyze_data_ref_access): For statements that access the same 
702         data-ref, check that they are not stores; for loads, check that there
703         is no store that access the same location.
704
705 2007-01-20  Roger Sayle  <roger@eyesopen.com>
706             Joseph Myers  <joseph@codesourcery.com>
707
708         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
709         (-fdump-rtl-stack): Likewise.
710         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
711         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
712         (-mpt-fixed): Use @minus{} for minus sign.
713         (Using Precompiled Headers): Fix under/overfull hbox.
714
715 2007-01-20  Jan Hubicka  <jh@suse.cz>
716
717         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
718         place within annotation.
719
720 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
721
722         * pa.c (output_move_double): Change array size of xoperands to 4.
723
724 2007-01-20  Jan Hubicka  <jh@suse.cz>
725
726         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
727         as temporary.
728
729 2007-01-19  Ian Lance Taylor  <iant@google.com>
730
731         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
732         constant to unsigned HOST_WIDE_INT before negating.
733
734 2007-01-19  Ian Lance Taylor  <iant@google.com>
735
736         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
737         to unsigned.
738         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
739         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
740         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
741         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
742         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
743         (realloc_vdef, realloc_vuse): Change second parameter to
744         unsigned.
745         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
746         fields to unsigned.
747         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
748         unsigned. Change x and lim locals to unsigned.
749         (realloc_vdef, realloc_vuse): Change num_elem parameter to
750         unsigned.
751         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
752         (copy_virtual_operands): Change i and n locals to unsigned.
753
754 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
755
756         PR middle-end/29335
757         * builtins.c (fold_builtin_1): Handle builtin fdim.
758
759 2007-01-20  Jan Hubicka  <jh@suse.cz>
760
761         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
762         * tree-ssa-structalias.c (compute_points_to_sets): Do call
763         init_alias_heapvars.
764         (init_alias_heapvars): Initialize only when not already initialized.
765         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
766
767 2007-01-19  Roger Sayle  <roger@eyesopen.com>
768
769         * common.opt (fsigned-zeros): New command line option.
770         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
771         of flag_unsafe_math_optimizations.
772         * opts.c (set_fast_math_flags): The -ffast-math command line option
773         implies -fno-signed-zeros.
774         (fast_math_flags_set_p): Likewise.
775
776         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
777         the documentation of -ffast-math appropriately.  Wrap long lines.
778
779 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
780
781         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
782         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
783
784 2007-01-19  Tomas Bily  <tbily@suse.cz>
785
786         * cgraphunit.c (cgraph_finalize_function): Updating of pid
787         * tree-profile.c:
788         (tree_init_ic_make_global_vars): New function
789         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
790         (tree_gen_ic_profiler): New function
791         (tree_gen_ic_func_profiler): New function
792         (tree_profiling): Added calling of tree_gen_ic_func_profiler
793         (tree_profile_hooks): Added hook for indirec/virtual calls
794         * value-prof.c (tree_find_values_to_profile): New case for
795         indirect calls
796         (tree_values_to_profile): Call for determining indirect/virtual
797         counters
798         (tree_indirect_call_to_profile): New function
799         (tree_ic_transform): New function
800         (tree_ic): New function
801         (find_func_by_pid): New function
802         (init_pid_map): New function
803         (tree_value_profile_transformations): Added check for
804         indirect/virtual call transformation
805         * value-prof.h (enum hist_type): New counter type for
806         indirect/virtual calls
807         (profile_hooks): Added new hook for profiling indirect/virtual
808         calls
809         * profile.c (instrument_values): New case for indirect/virtual
810         call added
811         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
812         (GCOV_COUNTER_V_INDIR): New counter type
813         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
814         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
815         call added
816         * cgraph.c: Definition of cgraph_max_pid
817         (cgraph_create_node): Default init of pid attribute
818         * cgraph.h: Declaration of cgraph_max_pid
819         (struct cgraph_node): Added pid attribute
820         * libgcov.c (__gcov_indirect_call_profiler): New function
821         (__gcov_one_value_profiler_body): New function
822         (__gcov_one_value_profiler): Body was moved to
823         __gcov_one_value_profiler_body and calls it
824
825 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
826         
827         * doc/gty.texi (Options): Document the mark_hook option to GTY.
828         * gengtype.c (write_types_data, write_func_for_structure,
829         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
830         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
831         (walk_type, write_func_for_structure): Generate the mark_hook if
832         needed.
833
834 2007-01-19  Jan Hubicka  <jh@suse.cz>
835
836         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
837         'early' argument take inlining mode argument specifying whether to
838         inline for size/speeed or all functions; add support for flattening;
839         improve dumpting.
840         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
841
842 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
843
844         PR c++/17947
845         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
846         the name of the declared identifier.
847                 
848 2007-01-19  Dirk Mueller  <dmueller@suse.de>
849
850         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
851         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
852         duplicate evaluation.
853
854 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
855
856         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
857         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
858         to XFMode, use acosxf2 and truncate result to requested mode.
859         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
860         (asin<mode>2): Similarly, with asin expanders.
861         (*fscalexf4_i387): Rename from *fscalexf4.
862         (expNcorexf3): New expander.
863         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
864         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
865         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
866         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
867         patterns for SSE math.
868         (exp10<mode>2): Similarly, with exp10 expanders.
869         (exp2<mode>2): Similarly, with exp2 expanders.
870         (expm1<mode>2): Similarly, with expm1 expanders.
871         (ldexp<mode>3): Similarly, with ldexp expanders.
872         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
873         Use gen_truncxf<mode>2_i387_noop to truncate result.
874
875 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
876
877         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
878         for ColdFire targets.
879
880 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
881             Richard Sandiford  <richard@codesourcery.com>
882
883         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
884         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
885         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
886         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
887         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
888         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
889
890 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
891
892         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
893         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
894         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
895         (*adddi_dilshr32_cf): New pattern.
896         (adddi3, subdi3): Remove first alternatives.
897
898 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
899
900         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
901         implemented using move.l, do not use its cc result for floating-point
902         comparisons.
903
904 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
905
906         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
907         from accepting 's' constraints if flag_pic.
908
909 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
910
911         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
912         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
913         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
914         condition from TARGET_68881 to TARGET_HARD_FLOAT.
915
916 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
917
918         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
919         ColdFire alternatives.
920         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
921         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
922         architectures that support ff1 instruction.
923
924 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
925             Julian Brown  <julian@codesourcery.com>
926
927         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
928         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
929         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
930         Add commentary.
931
932 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
933
934         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
935         and return a bool.
936         (output_move_const_into_data_reg, output_move_simode_const): Delete.
937         * config/m68k/m68k.c (const_method, const_int_cost): Take a
938         HOST_WIDE_INT instead of an rtx.
939         (m68k_rtx_costs): Update call accordingly.
940         (output_move_const_into_data_reg): Likewise.  Fix formatting.
941         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
942         Return a bool.
943         (output_move_simode_const): Update calls after above changes.
944         Rework to use automatic variables and predicates like MEM_P.
945         * config/m68k/m68k.md (pushexthisi_const): Update call to
946         valid_mov3q_const.
947
948 2007-01-19  Dirk Mueller  <dmueller@suse.de>
949
950         * tree-ssa-alias.c (perform_var_substitution): Fix typo
951         in dump_flags test.
952
953 2007-01-19  Richard Guenther  <rguenther@suse.de>
954
955         * builtins.c (expand_builtin_cexpi): Fall back to expanding
956         via cexp in case sincos is not available.
957
958 2007-01-19  Richard Guenther  <rguenther@suse.de>
959
960         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
961         * defaults.h (TARGET_HAS_SINCOS): Default to off.
962         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
963         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
964         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
965         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
966         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
967         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
968
969 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
970
971         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
972         New insn patterns.
973         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
974         (atan2xf3): Directly generate RTL pattern.
975         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
976         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
977         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
978         disable patterns for SSE math.
979         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
980         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
981         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
982         disable patterns for SSE math.
983
984 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
985
986         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
987
988 2007-01-18  Roger Sayle  <roger@eyesopen.com>
989
990         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
991         VIEW_CONVERT_EXPR to the same type as it's operand.
992
993 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
994
995         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
996         configured for long double 128.
997
998 2007-01-18  Mike Stump  <mrs@apple.com>
999
1000         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1001         error.
1002
1003 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1004
1005         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1006         type correct.
1007
1008 2007-01-18  Jan Hubicka  <jh@suse.cz>
1009
1010         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1011         greater than NUM_VOP_FREE_BUCKETS.
1012
1013 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1014
1015         * tree-ssa-structalias.c: Update comments.
1016         (ptabitmap_obstack): Removed.
1017         (pta_obstack): New.
1018         (oldpta_obstack): Ditto.
1019         (stats): Add a few members.
1020         (struct variable_info): Remove node, complex, address_taken, and
1021         indirect_target members. Add oldsolution member.
1022         (new_var_info): Do not initialize removed members.
1023         (constraint_expr_type): Remove INCLUDES.
1024         (constraint_graph): Add size, implicit_preds, rep,
1025         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1026         (FIRST_REF_NODE): New macro.
1027         (LAST_REF_NODE): Ditto.
1028         (FIRST_ADDR_NODE): Ditto.
1029         (find): New function.
1030         (unite): Ditto.
1031         (dump_constraint): Do not handle INCLUDES.
1032         (insert_into_complex): Do not insert duplicate constraints.
1033         (condense_varmap_nodes): Renamed and rewritten into ...
1034         (merge_node_constraints): This. Also fix bug in handling of
1035         offseted copy constraints.
1036         (clear_edges_for_node): No longer need to deal with preds at all,
1037         or removing associated preds/succs. 
1038         (merge_graph_nodes): Deal with indirect_cycles.
1039         Don't deal with predecessors.
1040         (add_implicit_graph_edge): New function.
1041         (add_pred_graph_edge): Ditto.
1042         (add_graph_edge): Don't deal with predecessors.
1043         (build_constraint_graph): Removed.
1044         (build_pred_graph): New function.
1045         (build_succ_graph): Ditto.
1046         (struct scc_info): Removed in_component. Added roots, dfs, and
1047         node_mapping. Remove visited_index, unification_queue.
1048         (scc_visit): Deal with union-find we do now.
1049         Deal with cycles with REF nodes.
1050         (collapse_nodes): Renamed and rewritten to ...
1051         (unify_nodes): This.
1052         (process_unification_queue): Removed.
1053         (topo_visit): Cleanup
1054         (do_da_constraint): Use find.
1055         (do_sd_constraint): Ditto.
1056         (do_ds_constraint): Ditto.
1057         (do_complex_constraint): Ditto.
1058         (init_scc_info): Update for removed and added members.
1059         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1060         (find_indirect_cycles): This.
1061         (equivalence_class): New variable.
1062         (label_visit): New function.
1063         (perform_variable_substitution): Rewritten.
1064         (free_var_substitution_info): New function.
1065         (find_equivalent_node): Ditto.
1066         (move_complex_constraints): Ditto.
1067         (eliminate_indirect_cycles): Ditto.
1068         (solve_graph): Only propagate changed bits.
1069         Use indirect cycle elimination.
1070         Use find.
1071         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1072         (tree_id_eq): Renamed to ...
1073         (tree_vi_eq): This. Update for member change
1074         (insert_id_for_tree): Renamed and rewritten to ...
1075         (insert_vi_for_tree): This.
1076         (lookup_id_for_tree): Renamed and rewritten to ...
1077         (lookup_vi_for_tree): This.
1078         (get_id_for_tree): Renamed and rewritten to ...
1079         (get_vi_for_tree): Ditto.
1080         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1081         (process_constraint): Don't handle INCLUDES.
1082         Remove special ADDRESSOF case.
1083         (find_func_aliases): Rewrite to use vi functions instead of id
1084         ones.
1085         (create_function_info_for): Ditto.
1086         (create_variable_info_for): Ditto.
1087         (intra_create_variable_infos): Ditto.
1088         (merge_smts_into): Ditto.
1089         (find_what_p_points_to): Ditto.
1090         (init_base_vars): Ditto.
1091         (init_alias_vars): Ditto.
1092         (remove_preds_and_fake_succs): New function.
1093         (dump_sa_points_to_info): Dump new stats.
1094         (dump_solution_for_var): Use find.
1095         (set_used_smts): Fix formatting.
1096         (compute_points_to_sets): Updated for new functions.
1097         (ipa_pta_execute): Ditto.
1098
1099 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1100             Richard Sandiford  <richard@codesourcery.com>
1101
1102         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1103         parallels.
1104         * calls.c (expand_call): If the return value is a PARALLEL,
1105         extract its first member.
1106         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1107         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1108         use FUNCTION_EXTRA_EPILOGUE.
1109         (m68k_function_value): Return a PARALLEL if the return value
1110         is of a pointer type.
1111         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1112         (FUNCTION_EXTRA_EPILOGUE): Remove.
1113         * config/m68k/m68k.md (D0_REG): New constant.
1114
1115 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1116
1117         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1118         output a NOP for empty epilogues.
1119
1120 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1121
1122         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1123         before function.  Extend register save check to include all
1124         registers, not just integer ones.
1125
1126 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1127
1128         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1129         (m68k_use_return_insn): ...this.
1130         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1131         * config/m68k/m68k.c (use_return_insn): Rename to...
1132         (m68k_use_return_insn): ...this.
1133         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1134         of USE_RETURN_INSN.
1135
1136 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1137
1138         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1139         nearest even, fix denormal rounding overflow.
1140
1141 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1142
1143         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1144         f%$smove and f%$move.
1145         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1146         (extendsfdf2_cf): Use fdmove instead of f%&move.
1147         (truncdfsf2_cf): Use fsmove instead of f%$smove.
1148         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1149
1150 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1151
1152         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1153         GPR<-GPR moves.
1154
1155 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1156
1157         * real.h (real_format): Add a canonical_nan_lsbs_set field.
1158         (coldfire_single_format): Declare.
1159         (coldfire_double_format): Likewise.
1160         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1161         of qnan_msb_set to determine the lower bits of a canonical
1162         NaN significand.
1163         (encode_ieee_double): Likewise.
1164         (encode_ieee_quad): Likewise.
1165         (ieee_single_format): Initialize canonical_nan_lsbs_set.
1166         (mips_single_format): Likewise.
1167         (ieee_double_format): Likewise.
1168         (mips_double_format): Likewise.
1169         (ieee_extended_motorola_format): Likewise.
1170         (ieee_extended_intel_96_format): Likewise.
1171         (ieee_extended_intel_128_format): Likewise.
1172         (ieee_extended_intel_96_round_53_format): Likewise.
1173         (ibm_extended_format): Likewise.
1174         (mips_extended_format): Likewise.
1175         (ieee_quad_format): Likewise.
1176         (mips_quad_format): Likewise.
1177         (vax_f_format): Likewise.
1178         (vax_d_format): Likewise.
1179         (vax_g_format): Likewise.
1180         (i370_single_format): Likewise.
1181         (i370_double_format): Likewise.
1182         (decimal_single_format): Likewise.
1183         (decimal_double_format): Likewise.
1184         (decimal_quad_format): Likewise.
1185         (c4x_single_format): Likewise.
1186         (c4x_extended_format): Likewise.
1187         (real_internal_format): Likewise.
1188         (coldfire_single_format): New real_format.
1189         (coldfire_double_format): Likewise.
1190         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1191         canonical_nan_lsbs_set.
1192         (pdp11_d_format): Likewise.
1193         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1194         if TARGET_COLDFIRE_CPU.
1195
1196 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1197
1198         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1199         (output_call, m68k_legitimize_call_address): Declare.
1200         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1201         parenthesees.  Add support for a 'W' constraint.
1202         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1203         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1204         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1205         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1206         variables.
1207         (override_options): Initialize them.  Do not set flag_no_function_cse
1208         for TARGET_ID_SHARED_LIBRARY.
1209         (m68k_output_pic_call): Delete.
1210         (m68k_legitimize_call_address): New function.
1211         (print_operand): Remove the %o prefix.  Handle the %p prefix.
1212         (output_call): New function.
1213         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
1214         target address from the GOT if symbolic jumps are not allowed.
1215         * config/m68k/m68k.md (call, general_operand): Do not set
1216         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
1217         Merge separate flag_pic and !flag_pic define_insns into...
1218         (*call, *call_value): ...these new patterns.  Match the address
1219         rather than the containing MEM and require it to be a call_operand.
1220         Use output_call to generate the asm template.
1221         * config/m68k/predicates.md (const_call_operand): New predicate.
1222         (call_operand): Likewise.
1223
1224 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1225
1226         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1227         (STACK_GROWS_DOWNWARD): Define to 1.
1228         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1229         comments, emphasizing that these values are only defaults.
1230         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1231         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1232
1233 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1234
1235         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1236
1237 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1238             Richard Sandiford  <richard@codesourcery.com>
1239
1240         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1241         (FP_REGNO_P): New macros.
1242         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1243         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1244         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1245         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1246
1247 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1248
1249         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1250         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1251         to $tmake_file.
1252         * config/m68k/t-floatlib: New file, extracting common code from...
1253         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1254         * config/m68k/t-uclinux: Here.
1255         * config/m68k/fpgnulib.c: Do not compile extendeed precision
1256         routines on ColdFire targets.
1257
1258 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1259
1260         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1261         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1262         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1263         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1264         explicitly set MOTOROLA to 1.
1265         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1266
1267 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1268             Nathan Sidwell  <nathan@codesourcery.com>
1269
1270         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1271         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1272         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1273         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1274         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1275         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1276         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1277         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1278         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1279         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1280         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1281         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1282         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1283         (DBX_REGISTER_NUMBER): Delete.
1284         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1285         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1286         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1287
1288 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1289
1290         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1291         * config/flat.h: New file.
1292         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1293         OBJECT_FORMAT_FLAT.
1294         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1295         indirect references for -msep-data or -mid-shared-library.
1296         Do not use PC-relative code addresses either.
1297         * config/m68k/m68k.c (override_options): Restrict -fPIC error
1298         to -mpcrel.
1299         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
1300         for shared libraries and crt1.o for executables.  Use crti.o and
1301         crtbegin.o.
1302         (ENDFILE_SPEC): Use crtend.o and crtn.o.
1303         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1304         Do not add -elf2flt or -shared-lib-id options here.
1305         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
1306         Pass -shared-lib-id if -mid-shared-library, taking the library
1307         identifier from -mshared-library-id if given, otherwise
1308         defaulting to 0.
1309         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1310         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1311         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1312         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1313         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1314         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1315         and crtend.o.
1316         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1317         jump sequence for ISA A and ISA A+.
1318         (PICJUMP): Likewise.
1319
1320 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1321
1322         * config.gcc (m68k-*-uclinux*): Base the port on the common
1323         and m68k GNU/Linux files rather than on the generic ELF ones.
1324         * config/m68k/uclinux.h (TARGET_VERSION): Override.
1325         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1326
1327 2007-01-18  Julian Brown  <julian@codesourcery.com>
1328             Richard Sandiford  <richard@codesourcery.com>
1329
1330         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1331         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1332         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1333         before redefining.
1334         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1335         80 unconditionally.
1336         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1337
1338 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1339
1340         * doc/install.texi: Document m68k-uclinuxoldabi.
1341         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1342         * config/m68k/uclinux-oldabi.h: New file, copied from
1343         config/m68k/uclinux.h.
1344
1345 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
1346
1347         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1348         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1349         (m32r_compute_frame_size): Use unsigned for regno.
1350         (m32r_reload_lr): Use gen_frame_mem.
1351         (pop): New.
1352         (m32r_output_function_epilogue): Don't output the function epilogue
1353         textually here.
1354         (m32r_expand_epilogue): New.
1355         (direct_return): Return false if the function has the interrupt
1356         attribute.
1357         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1358         * config/m32r/m32r.md (epilogue): New expander.
1359         (return_lr, return_rte): New insns.
1360         (return): Make it expander.
1361         (return_normal): New expander.
1362
1363 2007-01-18  Josh Conner  <jconner@apple.com>
1364
1365         PR target/30485
1366         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1367         support for UNLE, UNLT, UNGE, and UNGT.
1368
1369 2007-01-18  Jan Hubicka  <jh@suse.cz>
1370
1371         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1372         * tree-tailcall (add_virtual_phis): Likewise.
1373         (optimize_tail_call): Return TODOs.
1374         (execute_tail_calls): Return TODOs.
1375         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1376         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1377         changed.
1378         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1379         Cleanup cfg using TODO.
1380         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1381         * passes.c (execute_function_todo): When cleanup did something, remove
1382         unused locals.
1383         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1384         (make_edges): Don't cleanup_cfg.
1385
1386 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
1387
1388         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1389         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1390         operate on the top of stack.
1391         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1392         register handling of instructions that output to the second
1393         stack slot.
1394         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1395         (move_for_stack_reg): Special-case check for dead destination
1396         stack slot for constant load of 1.0 inside UNSPEC_TAN.
1397
1398         * config/i386/i386.md (UNSPEC_TAN): New constant.
1399         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1400         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1401         to correctly model move of constant 1.0 to top stack slot.
1402         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1403         (unnamed peephole2 pattern): Remove corresponding peephole2
1404         pattern that optimizes tan insn and loading of constant 1.0.
1405         (tanxf2): Use fptanxf4_i387.
1406         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1407         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
1408         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1409         disable patterns for SSE math.
1410         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1411         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1412
1413 2007-01-18  Dirk Mueller  <dmueller@suse.de>
1414             Richard Guenther <rguenther@suse.de>
1415
1416         PR diagnostic/8268
1417         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1418         * common.opt (Warray-bounds): Add new warning option.
1419         * c-opts.c (c_common_handle_option): Define -Warray-bounds
1420         if -Wall is given.
1421         * Makefile.in: make tree-vrp.o depend on toplev.h
1422         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1423         is enabled.
1424         (check_array_refs, check_array_bounds, check_array_ref): New.
1425
1426 2007-01-18  Jan Hubicka  <jh@suse.cz>
1427
1428         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1429         (execute_ssa_ccp): Return flags conditionally.
1430         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1431         changed.
1432         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1433
1434 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
1435
1436         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1437         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1438
1439         * Makefile.c (GTFILES): Add cfglayout.h.
1440         * gengtype.c (open_base_files): Likewise.
1441         * cfglayout.c (cfg_layout_function_footer,
1442         cfg_layout_function_header) Reindent to make gengtype happy.
1443         * cfglayout.h (cfg_layout_function_footer,
1444         cfg_layout_function_header): Add GTY(()) marker.
1445
1446         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1447
1448 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1449
1450         * genautomata.c (write_automata): Include xstrerror output in the
1451         error message if writing the DFA description file fails.
1452
1453 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1454
1455         * config/mips/mips-protos.h (mips_output_external): Make it
1456         return void.
1457         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1458         * config/mips/mips.c (irix_output_external_libcall): Likewise.
1459         (extern_list): Likewise.
1460         (extern_head): Likewise.
1461         (TARGET_ASM_FILE_END): Likewise.
1462         (mips_file_end): Likewise.
1463         (mips_output_external): Rewritten.
1464
1465 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1466
1467         * genpreds.c (write_insn_preds_c): Only write out the function
1468         body for regclass_for_constraint if we have register constraints.
1469
1470 2007-01-17  Tom Tromey  <tromey@redhat.com>
1471
1472         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1473         * doc/install.texi (Testing): Use sourceware.org.
1474         (Binaries): Likewise.
1475         (Specific): Likewise.
1476         * doc/contrib.texi (Contributors): Use sourceware.org.
1477
1478 2007-01-17  Anatoly Sokolov <aesok@post.ru>
1479
1480         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1481         (AVR_ENHANCED): Rename to ...
1482         (AVR_HAVE_MUL): ... new.
1483         (avr_enhanced_p): Rename to ...
1484         (avr_have_mul_p): ... new.
1485         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of 
1486         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1487         * config/avr/avr.c (avr_enhanced_p): Rename to ...
1488         (avr_have_mul_p): ... new.
1489         (base_arch_s): Rename 'enhanced' to 'have_mul'.
1490         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1491         'avr_enhanced_p' and 'enhanced'.
1492         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use 
1493         AVR_HAVE_MUL instead of AVR_ENHANCED.
1494         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1495         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3, 
1496         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL 
1497         instead of AVR_ENHANCED.
1498         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1499         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1500         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1501
1502 2007-01-17  Ian Lance Taylor  <iant@google.com>
1503
1504         * vec.h (VEC_reserve_exact): Define.
1505         (vec_gc_p_reserve_exact): Declare.
1506         (vec_gc_o_reserve_exact): Declare.
1507         (vec_heap_p_reserve_exact): Declare.
1508         (vec_heap_o_reserve_exact): Declare.
1509         (VEC_OP (T,A,reserve_exact)): New static inline function, three
1510         versions.
1511         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1512         negative parameter.
1513         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1514         (VEC_OP (T,A,copy)) [all versions]: Likewise.
1515         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1516         * vec.c (calculate_allocation): Add exact parameter.  Change all
1517         callers.
1518         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1519         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1520         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1521         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1522         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1523         (vec_heap_p_reserve_exact): New function.
1524         (vec_heap_o_reserve_exact): New function.
1525
1526 2007-01-17  Jan Hubicka  <jh@suse.cz>
1527
1528         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1529
1530 2007-01-17  Eric Christopher  <echristo@apple.com>
1531
1532         * config.gcc: Support core2 processor.
1533
1534 2007-01-16  Jan Hubicka  <jh@suse.cz>
1535
1536         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1537         the names, just unlink the chain so we don't crash on dangling pointers
1538         to dead SSA names.
1539
1540 2007-01-16  Jan Hubicka  <jh@suse.cz>
1541
1542         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1543         * tree-pass.h: Reorder to make IPA passes appear toegher.
1544         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1545         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1546         parameters, do not call early inliner.
1547         * ipa-inline.c: Update comments.  Include tree-flow.h
1548         (cgraph_decide_inlining): Do not compute inlining parameters.
1549         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1550         be called with function context set up.
1551         (pass_ipa_inline): Remove unreachable functions before pass.
1552         (cgraph_early_inlining): Simplify assuming to be called from the
1553         PM as local pass.
1554         (pass_early_inline): New pass.
1555         (cgraph_gate_ipa_early_inlining): New gate.
1556         (pass_ipa_early_inline): Turn into simple wrapper.
1557         (compute_inline_parameters): New function.
1558         (gate_inline_passes): New gate.
1559         (pass_inline_parameters): New pass.
1560         (apply_inline): Move here from tree-optimize.c
1561         (pass_apply_inline): New pass.
1562         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1563         transforming.
1564         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1565         doing them by hand.
1566         (tree_function_versioning): Do not allocate dummy struct function.
1567         * tree-inline.h (optimize_inline_calls): Update prototype.
1568         * tree-optimize.c (execute_fixup_cfg): Export.
1569         (pass_fixup_cfg): Remove
1570         (tree_rest_of_compilation): Do not apply inlines.
1571         * tree-flow.h (execute_fixup_cfg): Declare.
1572         * Makefile.in (gt-passes.c): New.
1573         * passes.c: Include gt-passes.h
1574         (init_optimization_passes): New passes.
1575         (nnodes, order): New static vars.
1576         (do_per_function_toporder): New function.
1577         (execute_one_pass): Dump current pass here.
1578         (execute_ipa_pass_list): Don't dump current pass here.
1579
1580 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
1581
1582         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1583         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1584
1585 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
1586
1587         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1588         (strong_alias): Define.
1589         (__gcc_qmul): Provide non-FMA for soft-float.
1590         (__gcc_qdiv): Same.
1591         (__gcc_qneg): New.
1592         (__gcc_qeq): New.
1593         (__gcc_qle): New.
1594         (__gcc_qge): New.
1595         (__gcc_qunord): New.
1596         (__gcc_stoq): New.
1597         (__gcc_dtoq): New.
1598         (__gcc_qtos): New.
1599         (__gcc_qtod): New.
1600         (__gcc_qtoi): New.
1601         (__gcc_qtou): New.
1602         (__gcc_itoq): New.
1603         (__gcc_utoq): New.
1604         (fmsub): New.
1605         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1606         soft-float functions.
1607         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1608         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1609         about long double soft float.
1610
1611 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
1612             Tehila Meyzels  <tehila@il.ibm.com>
1613
1614         * tree-vectorizer.h (is_pattern_stmt_p): New.
1615         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1616         formatting (tabs instead of spaces). Cleanup and clarify setting
1617         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1618         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1619         (vectorizable_type_demotion): Check that types are integral.
1620         (vectorizable_type_promotion): Likewise.
1621         (vectorizable_store): Fix typo.  Eliminate new-line at end of
1622         comments.
1623
1624 2007-01-16  Jan Hubicka  <jh@suse.cz>
1625
1626         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1627         cgraph edges too.
1628
1629 2007-01-15  Eric Christopher  <echristo@apple.com>
1630
1631         * ifcvt.c: Include vec.h, vecprim.h.
1632         (check_cond_move_block): New argument regs.
1633         Reorganize. Add registers used to regs.
1634         (cond_move_process_if_block): Use regs set above as
1635         loop bounds.
1636
1637 2007-01-15  Eric Christopher  <echristo@apple.com>
1638
1639         * config/darwin.h: Update copyright.
1640         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1641         (LINK_COMMAND_SPEC): Add -u.
1642         (LINK_SPEC): Fix umbrella for above.
1643
1644 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
1645
1646         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1647         glibc CVS.
1648
1649 2007-01-15  Tom Tromey  <tromey@redhat.com>
1650
1651         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1652         * doc/install.texi (Testing): Don't mention jacks.
1653         (Configuration): Document --enable-java-maintainer-mode.  Move
1654         --with-java-home to libgcj-specific section.  Document
1655         --with-ecj-jar.
1656         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1657
1658 2007-01-15  Jan Hubicka  <jh@suse.cz>
1659
1660         * tree-ssa-dce.c (DCE_TODOs): New.
1661         (propagate_necessity): Return if something changed.
1662         (eliminate_unnecessary_stmts): Likewise.
1663         (perform_tree_ssa_dce): Return TODO flags when needed.
1664         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1665
1666 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
1667
1668         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1669         (fyl2x_extend<mode>xf3_i387): New insn pattern.
1670         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1671         patterns using X87MODEF12 mode macro.  Extend operand 1
1672         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1673         (log10<mode>2): Ditto.
1674         (log2<mode>2): Ditto.
1675         (log1p<mode>2): Ditto.
1676         (logb<mode>2): Ditto.
1677         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1678         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1679         (*fxtractxf3_i387): Rename from *fxtractxf3.
1680         (fxtract_extend<mode>xf3_i387): New insn pattern.
1681         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1682         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1683         and gen_fyl2xxf3_i387().
1684
1685 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
1686
1687         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1688         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1689         get_loop_exit_edges.  Changed return type to bool.
1690         * cfgloop.h (fix_loop_placement): Declaration removed.
1691
1692 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1693
1694         * param.h (MIN_VECT_LOOP_BOUND): New.
1695         * params.def (MIN_VECT_LOOP_BOUND): New.
1696         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1697         argument - minimum threshold for number of iterations.
1698         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1699         argument to declaration.
1700         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1701         MIN_VECT_LOOP_BOUND.
1702         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1703         slpeel_tree_peel_loop_to_edge with additional argument.
1704         (vect_do_peeling_for_alignment): Likewise.
1705         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1706
1707 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1708
1709         PR target/30413
1710         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1711         operands of size 1.
1712
1713 2007-01-14  Jan Hubicka  <jh@suse.cz>
1714
1715         * tree-dfa.c (remove_referenced_var): New function.
1716         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1717         prune referenced vars list too.
1718         * tree-flow.h (remove_referenced_var): Declare.
1719
1720 2007-01-14  Jan Hubicka  <jh@suse.cz>
1721
1722         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1723         separately.
1724         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1725         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1726         * except.h (verify_eh_throw_table_statements): Kill prototype.
1727         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1728         into call exrepssions.
1729         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1730         eh trhow verifier.
1731         * tree-cfg.c: Include pointer-set.h.
1732         (verify_node_sharing): Work on pointer set.
1733         (verify_eh_throw_stmt_node): New.
1734         (verify_stmts): Use pointers sets, verify throw_stmt.
1735
1736 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1737
1738         * ipa-reference.c (analyze_function): Consider also addresses taken
1739         in phi nodes.
1740
1741 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1742
1743         * c-typeck.c (null_pointer_constant_p): Replace use of
1744         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1745         (build_c_cast): Likewise.
1746
1747 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1748
1749         * tree.h (force_fit_type_double): Remove unused final argument.
1750         * c-common.c (constant_expression_warning): Replace use of
1751         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1752         (convert_and_check): Likewise.
1753         (shorten_compare): Update call to force_fit_type_double.
1754         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1755         * convert.c (convert_to_pointer): Update call to
1756         force_fit_type_double.
1757         * fold-const.c (force_fit_type_double): Remove overflowed_const
1758         argument.
1759         (int_const_binop, fold_convert_const_int_from_int,
1760         fold_convert_const_int_from_real, fold_div_compare,
1761         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1762         fold_abs_const, fold_not_const): Remove the final argument from
1763         calls to force_fit_type_double.
1764
1765 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1766
1767         * configure.ac: Set insn to "nop" for spu-*-* also.
1768         * configure: Regenerate.
1769
1770 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
1771
1772         * builtins.def : Add BUILT_IN_FREE.
1773
1774 2007-01-12  Jan Hubicka  <jh@suse.cz>
1775
1776         PR tree-optimization/30443
1777         * tree-inline.c (tree_function_versioning): Do not optimize when
1778         cloning for inlining.
1779
1780 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
1781
1782         * doc/loop.texi: Document recording of loop exits.
1783         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1784         (update_single_exits_after_duplication,
1785         update_single_exit_for_duplicated_loop,
1786         update_single_exit_for_duplicated_loops): Removed.
1787         (duplicate_loop_to_header_edge): Do not call
1788         update_single_exits_after_duplication and
1789         update_single_exit_for_duplicated_loops.
1790         (loop_version): Do not update single_exit information.
1791         (fix_loop_structure): Use record_loop_exits instead of
1792         mark_single_exit_loops.
1793         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1794         the lists of loop exits.
1795         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1796         split_edge, merge_blocks): Update the lists of loop exits.
1797         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1798         loop_optimizer_init.
1799         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1800         of mark_single_exit_loops.
1801         (loop_optimizer_finalize): Call release_recorded_exits.
1802         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1803         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1804         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1805         update single exit information.
1806         * lambda-code.c (perfect_nestify): Ditto.
1807         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1808         (mark_single_exit_loops): Removed.
1809         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1810         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1811         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1812         functions.
1813         (get_loop_exit_edges, single_exit): Use recorded exit lists.
1814         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1815         (verify_loop_structure): Verify consistency of the exit lists.
1816         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
1817         (set_single_exit): Removed.
1818         * cfgloop.h (struct loop_exit): New function.
1819         (struct loop): single_exit_ field replaced by exits field.
1820         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1821         (struct loops): Added exits hash.
1822         (mark_single_exit_loops, set_single_exit): Declaration removed.
1823         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1824
1825 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1826
1827         * doc/invoke.texi: Avoid use of @headitem.
1828
1829 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1830
1831         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1832
1833 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1834
1835         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1836         list of targets.
1837
1838 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1839             Richard Sandiford  <richard@codesourcery.com>
1840
1841         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1842         * config/m68k/m68k.opt: Resort options.
1843         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1844
1845 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1846             Richard Sandiford  <richard@codesourcery.com>
1847
1848         * doc/invoke.texi: Document the macros that are defined by
1849         m68k's -mtune and -mhard-float options.
1850         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1851         (m68k_cpp_cpu_family): Likewise.
1852         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1853         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
1854         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
1855         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1856         (m68k_cpp_cpu_family): Likewise.
1857
1858 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1859
1860         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1861         macros besides mc68000 as tuning macros.  Use a switch statement
1862         to set them and mcpu32.
1863
1864 2007-01-12  Julian Brown  <julian@codesourcery.com>
1865
1866         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1867         (TARGET_68040_ONLY): Rename to...
1868         (TARGET_68040): ...this.
1869         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1870         * config/m68k/m68k.md: Likewise.
1871
1872 2007-01-12  Julian Brown  <julian@codesourcery.com>
1873             Nathan Sidwell  <nathan@codesourcery.com>
1874             Richard Sandiford  <richard@codesourcery.com>
1875
1876         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1877         the -mcpu= argument associated with the --with-cpu setting.
1878         Define M68K_DEFAULT_TUNE to the default -mtune= option,
1879         if different from the one implied by the -mcpu setting.
1880         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1881         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
1882         * doc/invoke.texi: Mention ColdFire in the introduction to the
1883         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
1884         -mno-div and -mhard-float options.  Make -m68881 a synonym for
1885         -mhard-float.  Document the previously-undocumented -m5206e,
1886         -m528x, -m5307 and -m5407 options.  Tweak the existing option
1887         documentation for consistency.
1888         * doc/install.texi: Mention new --with-cpu arguments.
1889         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1890         default CPU if neither -mcpu nor -march are specified.
1891         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1892         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1893         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
1894         using m68k_tune to decide between families that implement the
1895         same ISA.  Use m68k_tune to set __mcfv4e__.
1896         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1897         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1898         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1899         (FL_ISA_C, FL_ISA_MMU): New macros.
1900         (MASK_COLDFIRE): Delete.
1901         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1902         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1903         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1904         (TARGET_HARD_FLOAT): Do not define here.
1905         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1906         (TUNE_68000): New macro.
1907         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1908         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1909         (TUNE_CFV2): Redefine in terms of m68k_tune.
1910         (uarch_type, target_device, fpu_type): New enums.
1911         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1912         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1913         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1914         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1915         (FL_FOR_isa_c): New macros.
1916         (m68k_isa): New enum.
1917         (m68k_target_selection): New structure.
1918         (all_devices, all_isas, all_microarchs): New tables.
1919         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1920         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1921         (MASK_ALL_CPU_BITS): Delete.
1922         (m68k_find_selection): New function.
1923         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1924         Map the legacy target options to a combination of the new ones.
1925         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1926         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
1927         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1928         settings.
1929         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1930         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1931         properties.
1932         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1933         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1934         * config/m68k/m68k-devices.def: New file.
1935
1936 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1937             Nathan Sidwell  <nathan@codesourcery.com>
1938
1939         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1940         (SUBTARGET_EXTRA_SPECS): New macros.
1941         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1942         use %(asm_cpu_spec) instead.
1943         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1944         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1945         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1946         (EXTRA_SPECS): Rename to...
1947         (SUBTARGET_EXTRA_SPECS): ...this.
1948
1949 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1950             Richard Sandiford  <richard@codesourcery.com>
1951             Julian Brown  <julian@codesourcery.com>
1952
1953         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1954         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1955         to the configuration's default CPU.
1956         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1957         Remove default masks.
1958         (m680[012]0-*-*): Set the default with_cpu to the first part of
1959         the target name.
1960         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1961         (m68k*-*-linux): Extend the --with-cpu handling to...
1962         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
1963         and m68010.  Don't set target_cpu_default2.
1964         * doc/install.texi: Document --with-cpu for m68k.
1965         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1966         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1967         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1968         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1969         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1970         (ASM_SPEC): Remove use of %(asm_cpu_default).
1971         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1972         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1973         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
1974         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1975         __HAVE_FPU__ if TARGET_HARD_FLOAT.
1976         (TARGET_DEFAULT): Delete.
1977         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1978         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1979         (CPP_CPU_SPEC): Delete.
1980         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1981         the appropriate string.
1982         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1983         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1984         (ASM_SPEC): Don't use %(asm_default_spec).
1985         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1986         TARGET_DEFAULT and add MASK_68881.
1987         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1988
1989 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1990
1991         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1992         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1993         MASK_68010 alongside MASK_68020.
1994         * doc/invoke.texi: Document -m68010.
1995         * config/m68k/m68k.opt (m68010): New.
1996         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1997         if TUNE_68010.
1998         (TUNE_68010): New macro.
1999         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2000         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2001         (M68K_CPU_m68332): Add MASK_68010.
2002         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2003         fallback definition.
2004         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2005         defines.
2006         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2007         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2008         to all entries that use MASK_68020.
2009         (output_move_simode_const, output_move_himode, output_move_qimode)
2010         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2011         instead of TARGET_68020 to select clr behavior.  Remove comment
2012         about there being no TARGET_68010.
2013         * config/m68k/m68k.md: Likewise throughout.
2014
2015 2007-01-12  Julian Brown  <julian@codesourcery.com>
2016
2017         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2018         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2019         * config/m68k/m68k.md: Likewise.
2020
2021 2007-01-12  Julian Brown  <julian@codesourcery.com>
2022
2023         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2024         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2025
2026 2007-01-12  Julian Brown  <julian@codesourcery.com>
2027
2028         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2029         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2030         * config/m68k/m68k.md: Likewise.
2031
2032 2007-01-12  Julian Brown  <julian@codesourcery.com>
2033             Richard Sandiford  <richard@codesourcery.com>
2034
2035         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2036         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2037         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2038         TARGET_CPU32.
2039         (TARGET_CPU32): Rename to...
2040         (TUNE_CPU32): ...this.
2041         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2042         (TUNE_CFV2): New macros.
2043         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2044         remove conditions that are implied by TARGET_68020.
2045         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2046         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2047         (m68k_output_function_epilogue): Likewise.
2048         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2049         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2050         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2051         non-68000 timings.  Refactor multiplication and division costs.
2052         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2053         TUNE_CPU32 instead of TARGET_CPU32.
2054         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2055         and TUNE_68060 instead of TARGET_68060.
2056         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2057         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2058         TARGET_CPU32.
2059         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2060         && !TARGET_COLDFIRE" to choose between moveq and clr.
2061         Likewise in the unnamed movsf pattern.
2062         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2063         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2064         ashiftrt pattern.
2065
2066 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2067
2068         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2069         of tabbing before backslashes.
2070
2071 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2072
2073         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2074         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2075         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2076
2077 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2078
2079         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2080         find_interesting_uses_cond.
2081         (find_interesting_uses_cond): Use extract_cond_operands.
2082         (rewrite_use_compare): Use extract_cond_operands and
2083         force_gimple_operand_bsi.  Do not call update_stmt.
2084         (determine_use_iv_cost_condition): Use extract_cond_operands.
2085         Return cheaper of using original bound and changing the exit bound.
2086
2087 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2088
2089         PR tree-optimization/29516
2090         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2091         most_expensive_mult_to_index, addr_to_parts,
2092         create_mem_ref, maybe_fold_tmr): Make the type of
2093         fields of TARGET_MEM_REF sizetype.
2094         (move_fixed_address_to_symbol, move_pointer_to_base):
2095         New functions.
2096         * tree.def (TARGET_MEM_REF): Add comment on types of
2097         the operands.
2098
2099 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
2100
2101         * c-common.c (vector_types_convertible_p): Treat opaque types as
2102         always convertible if they have the same size, but not otherwise.
2103
2104 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
2105
2106         * ifcvt.c (struct noce_if_info): Add comments to the fields.
2107         Remove the b_unconditional field.
2108         (noce_try_sign_mask): Do not look at b_unconditional.
2109         (noce_process_if_block): Do not use merge_if_blocks.  Update
2110         the CFG here.  Do not set b_unconditional.
2111         (cond_move_process_if_block): Likewise.
2112         (find_cond_trap): Likewise.
2113         (check_cond_move_block): Require simple jump insns at the end
2114         of the basic block.
2115
2116 2007-01-11  Jan Hubicka  <jh@suse.cz>
2117
2118         PR tree-optimization/1046
2119         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2120         when alias info is not ready.
2121         (pass_tail_recursion): Do not require aliasing.
2122         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2123         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2124         * tree-ssa-copy.c (pass_copy_prop): Likewise.
2125         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2126         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2127         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2128         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2129         before inlining.
2130         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2131         When aliasing is not build, mark statement as volatile.
2132
2133 2007-01-11  Tom Tromey  <tromey@redhat.com>
2134
2135         PR preprocessor/15185, PR preprocessor/20989:
2136         * doc/cppopts.texi <-MT>: Update description of algorithm for
2137         computing default target.
2138         <-M, -MD>: Reword "basename" text.
2139
2140 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2141
2142         * builtins.c (expand_builtin_pow, expand_builtin_powi,
2143         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2144         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2145         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2146         fold_builtin_bswap, real_constp, fold_builtin_pow,
2147         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2148         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2149         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2150         * convert.c (convert_to_pointer): Likewise.
2151         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2152         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2153         const_binop, fold_convert_const_int_from_int,
2154         fold_convert_const_int_from_real,
2155         fold_convert_const_real_from_real, sign_bit_p,
2156         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2157         fold_sign_changed_comparison, fold_unary, fold_comparison,
2158         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2159         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2160         * print-tree.c (print_node_brief, print_node): Likewise.
2161         * stor-layout.c (place_field, layout_type): Likewise.
2162         * tree-chrec.c (keep_cast): Likewise.
2163         * tree.c (build_vector, build_real, build_real_from_int_cst,
2164         build_complex): Likewise.
2165
2166 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2167
2168         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
2169         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2170
2171 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
2172
2173         * configure.ac (strict1_warn): Rename to strict_warn.
2174         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2175         cc_set_by_configure, quoted_cc_set_by_configure,
2176         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2177         all_boot_languages, all_stagestuff): Remove.
2178         (target_list): Remove bootstrap targets.
2179         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2180         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2181         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2182         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2183         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2184         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2185         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2186         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2187         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2188         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2189         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2190         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2191         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2192         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2193         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2194         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2195         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2196         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2197         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2198         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2199         (mostlyclean): Adjust.
2200         (clean, distclean): Don't mention bootstrap stuff.
2201         * configure: Regenerate.
2202         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2203         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2204         treelang/config-lang.in (stagestuff): Remove.
2205         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2206         stageprofile, stagefeedback, stagestuff): Remove mention.
2207
2208 2007-01-11  Nick Clifton  <nickc@redhat.com>
2209
2210         * config/mcore/predicates.md (mcore_general_movesrc_operand):
2211         Accept CONSTs.
2212         (mcore_general_movdst_operand): Do not accept CONST_INTs.
2213         (mcore_arith_K_S_operand): Run the test for the S constraint not
2214         the test for the M constraint.
2215         (mcore_addsub_operand): Do not accept integer values that are
2216         larger than 32 bits.
2217         * config/mcore/mcore.md: Remove unused constraints from split.
2218         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2219         (addsi3): Likewise.
2220         (allocate_stack): Likewise.
2221         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2222         operands to 32 bits.
2223         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2224         INTVAL.
2225         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2226         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2227         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2228         mcore_output_bclri, mcore_output_andn, output_inline_const,
2229         mcore_output_move, mcore_output_movedouble): Likewise.
2230         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2231         (output_inline_const): Likewise.
2232         (output_inline_const): Fix format strings used in sprintf
2233         statements.
2234         * config/mcore/mcore-protos.h: Update prototypes for changed
2235         functions in mcore.c.
2236         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2237         HOST_WIDE_INT and not int.
2238         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2239         CONST_OK_FOR_N): Likewise.
2240         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2241         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2242         an INTVAL.
2243
2244 2007-01-10  Jan Hubicka  <jh@suse.cz>
2245
2246         * tree-vrp.c (remove_range_assertions): Release defs.
2247         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2248         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2249
2250 2007-01-10  Paul Brook  <paul@codesourcery.com>
2251
2252         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2253         * config/arm/arm.md (mulsi3subsi): New insn.
2254
2255 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
2256
2257         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2258         (tree_transform_and_unroll_loop): New.
2259         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2260         Declare.
2261
2262 2007-01-10  Robert Kennedy <jimbob@google.com>
2263
2264         * fold-const.c (fold_comparison): Fold comparisons like (x *
2265         1000 < 0) to (x < 0).
2266
2267 2007-01-10  Ian Lance Taylor  <iant@google.com>
2268
2269         * tree-pretty-print.c (dump_generic_node): Print parentheses when
2270         operands have the same priority.
2271
2272 2007-01-10  Tom Tromey  <tromey@redhat.com>
2273
2274         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2275         (fold_binary): Likewise.
2276         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2277         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2278         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2279         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2280         * langhooks.h (struct lang_hooks): Removed field
2281         'can_use_bit_fields_p'.
2282
2283 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
2284
2285         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2286
2287 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
2288
2289         * function.c (get_last_funcdef_no): New function.
2290         * function.h (get_last_funcdef_no): Declare.
2291         * tree-inline.c (initialize_cfun): Add initialization.
2292         (tree_function_versioning): Cleanup.
2293
2294 2007-01-10  Jan Hubicka  <jh@suse.cz>
2295
2296         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2297         PHIs.
2298
2299 2007-01-10  Sa Liu  <saliu@de.ibm.com>
2300             Ben Elliston  <bje@au.ibm.com>
2301
2302         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2303         for SAVE_NONLOCAL and Pmode for any other save level.
2304         * spu-protos.h (spu_restore_stack_block): Declare.
2305         * spu.md (save_stack_block): Remove.
2306         (restore_stack_block): Call spu_restore_stack_block.
2307         * spu.c (spu_restore_stack_block): New function.
2308         (spu_expand_epilogue): Remove old comment.
2309
2310 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
2311
2312         PR tree-optimization/30322
2313         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2314         (cand_value_at): Return the value as aff_tree.
2315         (may_eliminate_iv): Convert the bound from aff_tree to tree.
2316         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2317         aff_combination_mult): New functions.
2318         (aff_combination_add): Use aff_combination_add_cst.
2319         (aff_combination_convert): Allow conversions to a wider type.
2320         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2321         * tree-affine.h (aff_combination_mult): Declare.
2322
2323 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
2324
2325         * doc/tm.texi: Update documentation to reflect reality of exec
2326         and start file search behaviours. Update copyright year.
2327         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2328         header file directories.
2329
2330 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
2331
2332         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2333         (*cosxf2): Rename to cosxf2_i387.
2334         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
2335         using X87MODEF12 mode macro. Rename patterns to
2336         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
2337         for SSE math.
2338         (*cosdf2, *cossf2): Ditto.
2339         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
2340         to match extended input operands.
2341         (sincos<mode>3): New expander.
2342         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2343         insn patterns and corresponding splitters.
2344
2345 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
2346
2347         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2348         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2349
2350 2007-01-09  Nicolas Pitre  <nico@cam.org>
2351
2352         PR target/30173
2353         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2354
2355 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2356
2357         * target.h (struct gcc_target): New field library_rtti_comdat.
2358         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2359         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2360         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2361         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2362
2363 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2364
2365         * doc/invoke.texi (Optimize Options): Correct description of -O0.
2366
2367 2007-01-08  Richard Guenther  <rguenther@suse.de>
2368
2369         * tree.h (force_fit_type_double): Export.
2370         (force_fit_type): Remove.
2371         * fold-const.c (force_fit_type_double): New function.
2372         (force_fit_type): Remove.
2373         (int_const_binop): Use it.
2374         (fold_convert_const_int_from_int): Likewise.
2375         (fold_convert_const_int_from_real): Likewise.
2376         (fold_div_compare): Likewise.
2377         (fold_sign_changed_comparison): Likewise.
2378         (fold_unary): Likewise.
2379         (fold_negate_const): Likewise.
2380         (fold_abs_const): Likewise.
2381         (fold_not_const): Likewise.
2382         * c-common.c (shorten_compare): Use force_fit_type_double.
2383         * convert.c (convert_to_pointer): Likewise.
2384
2385 2007-01-08  Richard Guenther  <rguenther@suse.de>
2386
2387         * tree.h (build_int_cst_wide_type): Export.
2388         * tree.c (build_int_cst_wide_type): New function.
2389         (build_int_cst_wide): Fix comment.
2390         * builtins.c (fold_builtin_object_size): Use build_int_cst
2391         to build -1 or 0 of the correct type.  Use fit_double_type
2392         to check for overflow.
2393         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2394         to build the mask.
2395         (decode_field_reference): Likewise.
2396         (all_ones_mask_p): Likewise.
2397         (native_interpret_int): Use build_int_cst_wide_type.
2398         (fold_binary): Use build_int_cst_type to build an all-ones
2399         value.
2400         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2401
2402 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
2403
2404         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2405
2406 2007-01-09  Ben Elliston  <bje@au.ibm.com>
2407
2408         * genautomata.c (STATS_OPTION): New option.
2409         (stats_flag): New flag.
2410         (gen_automata_option): Handle it.
2411         (initiate_automaton_gen): Ditto.
2412         (write_automata): Output statistics only if stats_flag is
2413         set. Likewise, output time statistics only if time_flag is set.
2414         * doc/md.texi (Processor pipeline description): Document new flag.
2415
2416 2007-01-08  Richard Guenther  <rguenther@suse.de>
2417
2418         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2419         * tree.c (build_int_cst_type): Likewise.
2420         (size_in_bytes): Don't call force_fit_type on the result.
2421         (int_fits_type_p): Use fit_double_type.
2422         * fold-const.c (fit_double_type): New function.
2423         (force_fit_type): Use it.
2424         * tree.h (fit_double_type): Export.
2425
2426 2007-01-08  Jan Hubicka  <jh@suse.cz>
2427
2428         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2429         * ipa.c (function_and_variable_visibility): Fix return type.
2430
2431 2007-01-08  Richard Guenther  <rguenther@suse.de>
2432
2433         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2434         of offset to build the index.
2435         * tree-pretty-print.c (dump_generic_node): Don't build negated
2436         const just for printing.
2437         * c-pretty-print.c (pp_c_integer_constant): Likewise.
2438         * builtins.c (fold_builtin_int_roundingfn): Check if result
2439         fits the type by using force_fit_type and comparing the result.
2440         * predict.c (predict_loops): Use compare_tree_int for comparison.
2441         * tree.c (build_int_cst): Fall back to integer_type_node for
2442         NULL_TREE type.
2443         (build_int_cst_wide): Assert type is non-null.
2444
2445 2007-01-08  Roberto Costa  <roberto.costa@st.com>
2446
2447         * tree-vrp.c (extract_range_from_cond_expr): New.
2448         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2449         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2450         as expressions.
2451         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2452         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2453         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2454         used as expressions.
2455         * tree-object-size.c (cond_expr_object_size): New.
2456         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2457
2458 2007-01-08  Jan Hubicka  <jh@suse.cz>
2459
2460         * tree-ssa-forwprop.c (forward_propagate_into_cond,
2461         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2462         statement.
2463
2464 2007-01-08  Richard Guenther  <rguenther@suse.de>
2465
2466         PR tree-optimization/23603
2467         * tree-vrp.c (set_value_range_to_truthvalue): New function.
2468         (extract_range_from_binary): Fall back to truthvalue instead of
2469         varying for TRUTH_*_EXPR.
2470         (extract_range_from_comparison): Fall back to truthvalue instead of
2471         varying.
2472         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2473         if all visited PHI values were constant.
2474
2475 2007-01-08  Jan Hubicka  <jh@suse.cz>
2476
2477         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2478         (cgraph_analyze_function): break out from ...
2479         (cgraph_finalize_compilation_unit): ... here.
2480         (cgraph_expand_function): Remove forgoten commented out line.
2481         (cgraph_optimize): Analyze functions.
2482
2483 2007-01-08  Jan Hubicka  <jh@suse.cz>
2484
2485         * tree-pas.h (TODO_remove_function): New flag.
2486         (TODO_update*): Renumber.
2487         (pass_ipa_increase_alignment,
2488         pass_ipa_function_and_variable_visibility): New passes.
2489         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2490         (cgraph_function_and_variable_visibility): Move to ipa.c
2491         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2492         cgraph_increase_alignment.
2493         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2494         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2495         calling tree-inline.
2496         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2497         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2498         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2499         (gate_increase_alignment): New function.
2500         (pass_ipa_increase_alignment): New pass.
2501         * ipa.c: Inline tree-pass.h and timevar.h
2502         (function_and_variable_visibility): Move here from cgraphunit.c
2503         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2504         * passes.c (init_optimization_passes): Add the two new passes.
2505         (execute_todo): Handle cgraph_remove_functions.
2506
2507 2007-01-08  Nick Clifton  <nickc@redhat.com>
2508
2509         * config/frv/predicates.md (reg_or_0_operand): Accept
2510         CONST_DOUBLEs.
2511
2512 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
2513
2514         * config/bfin/rtems.h, config/bfin/t-rtems: New.
2515         * config.gcc: Add bfin*-rtems*.
2516
2517 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2518
2519         * c.opt: Add -flax-vector-conversions.
2520         * c-typeck.c (convert_for_assignment): Pass flag to
2521         vector_types_convertible_p to allow emission of note.
2522         (digest_init): Likewise.
2523         * c-opts.c: Handle -flax-vector-conversions.
2524         * c-common.c (flag_lax_vector_conversions): New.
2525         (vector_types_convertible_p): Unless -flax-vector conversions
2526         has been passed, disallow conversions between vectors with
2527         differing numbers of subparts and/or element types.  If such
2528         a conversion is disallowed, possibly emit a note on the first
2529         occasion only to inform the user of -flax-vector-conversions.
2530         The new last argument specifies this.
2531         * c-common.h (flag_lax_vector_conversions): New.
2532         (vector_types_convertible_p): Add extra argument.
2533         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2534         char_type_node for V*QI type vectors.
2535         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2536         Update to satisfy new typechecking rules.
2537         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2538         C and C++ variants.
2539         * doc/invoke.texi (C Dialect Options): Document
2540         -flax-vector-conversions.
2541
2542 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2543
2544         PR tree-optimization/29877
2545         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2546         a register variable on the RHS to not be replaceable.
2547
2548 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
2549         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2550         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2551         const_simm15, const_pow2, const_npow2): Added.
2552         * config/score/misc.md (insv, extv, extzv, movmemsi,
2553         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2554         * config/score/score.c (score_address_cost, score_select_cc_mode):
2555         Added.
2556         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2557         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2558         Update score_print_operand.
2559         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2560         Adjust register allocate order and update some macro define.
2561         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2562         mdx_block_move_straight, mdx_block_move_loop_head,
2563         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2564         mdx_block_move): Added.
2565         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2566         mdp_limm): Updated and fix some bug and typo.
2567         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2568         (movsf, movdf, doloop_end): Added.
2569
2570 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
2571
2572         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2573         config/arm/thumb2.md: Fix comment typos.
2574         * doc/extend.texi: Fix a typo.
2575
2576 2007-01-07  Eric Christopher  <echristo@apple.com>
2577
2578         * configure.ac: Check for __stack_chk_fail for darwin.
2579         * configure: Regenerate.
2580
2581 2007-01-07  Richard Guenther  <rguenther@suse.de>
2582
2583         * tree-vrp.c (extract_range_from_assert): CSE calls to
2584         compare_values where possible.
2585         (extract_range_from_unary_expr): Likewise.
2586
2587 2007-01-07  Anatoly Sokolov <aesok@post.ru>
2588
2589         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2590         * config/avr/avr.c (call_insn_operand): Delete function.
2591         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2592         instead of register number. Use predicates.
2593         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2594         call_insn_operand): Add.
2595
2596 2007-01-06  Jan Hubicka  <jh@suse.cz>
2597
2598         * tree-pass.h (pass_build_cgraph_edges): Declare.
2599         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2600         (visited_nodes): Remove.
2601         (cgraph_create_edges): Move to cgraphbuild.c; rename to
2602         build_cgrpah_edges; make visited_nodes local.
2603         (cgraph_process_new_functions): DO not call initialize_inline_failed.
2604         (record_references_in_initializer): Move to cgraphbuild.c
2605         (initialize_inline_failed, rebuild_cgraph_edges,
2606         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2607         (verify_cgraph_node): Make visited_nodes local.
2608         (cgraph_analyze_function): Do not call cgraph_create_edges and
2609         initialize_inline_failed.
2610         (cgraph_expand_function): Do not call cgraph_lower_function;
2611         assert that function is already lowered.
2612         * Makefile.in (cgraphbuild.o): New.
2613         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2614         at the end of lowering passes.
2615
2616 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
2617
2618         * ifcvt.c (cond_move_convert_if_block): New function, code
2619         factored out from...
2620         (cond_move_process_if_block): ...here.  Call the new function
2621         on the THEN and ELSE blocks.
2622         (merge_if_block): Do not copy global_live_at_end, merge_blocks
2623         already takes care of this.
2624
2625 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2626
2627         PR c/19978
2628         * tree.h (TREE_OVERFLOW_P): New.
2629         * c-typeck.c (parser_build_unary_op): Warn only if result
2630         overflowed and operands did not.
2631         (parser_build_binary_op): Likewise.
2632         (convert_for_assignment): Remove redundant overflow_warning.
2633         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2634
2635 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2636
2637         * c-typeck.c (store_init_value): Split over two lines to follow
2638         the GNU coding style.
2639
2640 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
2641
2642         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2643         __GXX_EXPERIMENTAL_CXX0X__.
2644         * doc/cpp.texi: Same.
2645
2646 2007-01-05  Richard Guenther  <rguenther@suse.de>
2647
2648         PR middle-end/27826
2649         * tree.c (get_narrower): Do not construct COMPONENT_REFs
2650         with mismatched types.  Instead explicitly build a
2651         conversion NOP_EXPR.
2652
2653 2007-01-05  Ian Lance Taylor  <iant@google.com>
2654
2655         * c-common.c (decl_with_nonnull_addr_p): New function.
2656         (c_common_truthvalue_conversion): Call it.
2657         * c-typeck.c (build_binary_op): Likewise.
2658         * c-common.h (decl_with_nonnull_addr_p): Declare.
2659
2660 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
2661
2662         PR c/30360
2663         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2664         denom against 0.0.
2665
2666 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
2667
2668         * doc/install.texi (Final install): Document the fact that
2669         the GNAT runtime should not be stripped.
2670
2671 2007-01-04  Jan Hubicka  <jh@suse.cz>
2672
2673         * tree-inline.c (fold_marked_statements): Update operand caches
2674         and EH after folding
2675
2676 2007-01-04  Ian Lance Taylor  <iant@google.com>
2677
2678         * c-common.c (check_function_nonnull): Whitespace fix.
2679
2680 2007-01-04  Jan Hubicka  <jh@suse.cz>
2681
2682         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2683         comitted older version of patch.
2684         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2685         function, celanup cfg and collect garbage.
2686
2687 2007-01-04  Mike Stump  <mrs@apple.com>
2688
2689         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2690         (clean): Likewise.
2691
2692 2007-01-04  Eric Christopher  <echristo@apple.com>
2693
2694         * libgcc2.c (__bswapsi2): Use SItype.
2695         (__bswapdi2): Use DItype.
2696         * libgcc2.h: Update for above.
2697
2698 2007-01-04  Paul Brook  <paul@codesourcery.com>
2699
2700         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2701         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2702         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2703         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2704         umulsi3_highpart): Make conditional on !arm_arch6.
2705         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2706         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2707         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2708         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2709         umulsi3_highpart_v6): New insns.
2710
2711 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2712
2713         * fold-const.c (fold_convert): When casting an expression to void,
2714         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2715         have a type.  Instead of attempting to build a NOP_EXPR, return
2716         these "special" trees directly.
2717
2718 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2719
2720         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2721         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2722         rs6000_cost->fp not 0 as adjustment for outer NEG.
2723
2724 2007-01-04  Jan Hubicka  <jh@suse.cz>
2725
2726         * cgraph.c (cgraph_release_function_body): New function.
2727         (cgraph_remove_node): Use it.
2728         * cgraph.h (cgraph_release_function_body): Declare.
2729         * cgraphunit.c (cgraph_expand_function): Use it.
2730         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2731         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2732         is initialized and while compilation of other function is running.
2733         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2734         statement CFG annotations from here to ...
2735         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2736
2737 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2738
2739         * cfgloop.h (enum li_flags): Make the constants powers of two.
2740
2741 2007-01-04  Jan Hubicka  <jh@suse.cz>
2742
2743         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2744         set.
2745         (fold_marked_statements): New function.
2746         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2747         * tree-inline.h (copy_body_data): Add statements_to_fold.
2748
2749 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2750
2751         * config.gcc: Mention libgcc/config.host.
2752         * Makefile.in: Update comments mentioning libgcc.
2753         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2754         (all.cross, start.encap, rest.encap, rest.cross): Update
2755         dependencies for libgcc move.
2756         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2757         (clean-target, clean-target-libgcc): Delete.
2758         (srcdirify, GCC_EXTRA_PARTS): New macros.
2759         (libgcc-support, libgcc.mvars): New rules.
2760         (distclean): Remove mention of mklibgcc.
2761         (install): Don't reference INSTALL_LIBGCC.
2762         (install-common): Don't reference EXTRA_PARTS.
2763         (install-libgcc, install-multilib): Delete rules.
2764         * mklibgcc.in: Delete file.
2765         * doc/configfiles.texi: Don't mention mklibgcc.
2766
2767         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2768         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2769         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2770         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2771         (mklibgcc, ldblspecs): Likewise.
2772
2773         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2774         (posixpre.def): Use $(T).
2775         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2776         (SHLIB_SRC, SHLIB_INSTALL): Delete.
2777         (SHLIB_LINK): Make dummy.
2778         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2779
2780         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2781
2782         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2783         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2784         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2785         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2786
2787         * configure.ac (all_outputs): Remove mklibgcc.
2788         * configure: Regenerated.
2789
2790 2007-01-03  Josh Conner  <jconner@apple.com>
2791
2792         PR middle-end/29683
2793         * calls.c (compute_argument_addresses): Set stack and stack_slot
2794         for partial args, too.
2795         (store_one_arg): Use locate.size.constant for the size when
2796         generating a save_area.
2797
2798 2007-01-03  Robert Kennedy <jimbob@google.com>
2799
2800         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2801         whose definitions are deleted due to basic block merging.
2802
2803 2007-01-03  Paul Brook  <paul@codesourcery.com>
2804
2805         PR target/16634
2806         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2807         functions.
2808         (use_return_insn): Return 0 for Thumb interrupt functions.
2809         (print_multi_reg): Add rfe argument for IRQ returns.
2810         (arm_output_epilogue): Pop interrupt return address directly into PC.
2811         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2812
2813 2007-01-03  Paul Brook  <paul@codesourcery.com>
2814
2815         Merge from sourcerygxx-4_1.
2816         * config/arm/thumb2.md: New file.
2817         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2818         Thumb-2.
2819         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2820         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2821         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2822         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2823         tables.
2824         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2825         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2826         comptibility.
2827         * config/arm/ieee754-sf.S: Ditto.
2828         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2829         (thumb1_base_register_rtx_p): ... to this.
2830         (thumb_index_register_rtx_p): Rename...
2831         (thumb1_index_register_rtx_p): ... to this.
2832         (thumb_output_function_prologue): Rename...
2833         (thumb1_output_function_prologue): ... to this.
2834         (thumb_legitimate_address_p): Rename...
2835         (thumb1_legitimate_address_p): ... to this.
2836         (thumb_rtx_costs): Rename...
2837         (thumb1_rtx_costs): ... to this.
2838         (thumb_compute_save_reg_mask): Rename...
2839         (thumb1_compute_save_reg_mask): ... to this.
2840         (thumb_final_prescan_insn): Rename...
2841         (thumb1_final_prescan_insn): ... to this.
2842         (thumb_expand_epilogue): Rename...
2843         (thumb1_expand_epilogue): ... to this.
2844         (arm_unwind_emit_stm): Rename...
2845         (arm_unwind_emit_sequence): ... to this.
2846         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2847         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2848         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2849         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2850         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2851         thumb2_output_casesi): New functions.
2852         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2853         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2854         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2855         THUMB2_WORK_REGS): Define.
2856         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2857         arm_condexec_mask, arm_condexec_masklen)): New variables.
2858         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2859         (arm_override_options): Check new CPU capabilities.
2860         Set new architecture flag variables.
2861         (arm_isr_value): Handle v7m interrupt functions.
2862         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
2863         Thumb-2.
2864         (const_ok_for_arm): Handle Thumb-2 constants.
2865         (arm_gen_constant): Ditto.  Use movw when available.
2866         (arm_function_ok_for_sibcall): Return false for v7m interrupt
2867         functions.
2868         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2869         (thumb_find_work_register, arm_load_pic_register,
2870         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2871         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2872         print_multi_reg, output_mov_long_double_fpa_from_arm,
2873         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2874         output_mov_double_fpa_from_arm, output_move_double,
2875         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2876         output_return_instruction, arm_output_function_prologue,
2877         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2878         arm_output_mi_thunk, thumb_set_return_address): Ditto.
2879         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
2880         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2881         (arithmetic_instr, shift_op): Use arm_shift_nmem.
2882         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
2883         '!' and 'L'.
2884         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2885         (thumb_expand_prologue): Use thumb_set_frame_pointer.
2886         (arm_file_start): Output directive for unified syntax.
2887         (arm_unwind_emit_set): Handle stack alignment instruction.
2888         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2889         Add v6t2, v7, v7a, v7r and v7m.
2890         (RETLDM): Add Thumb-2 code.
2891         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2892         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2893         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2894         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2895         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2896         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2897         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2898         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2899         ADJUST_INSN_LENGTH): Define.
2900         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2901         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2902         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2903         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2904         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2905         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2906         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2907         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2908         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2909         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2910         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2911         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2912         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2913         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2914         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2915         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2916         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2917         cortex-m3.
2918         * config/arm/arm-tune.md: Regenerate.
2919         * config/arm/arm-protos.h: Update prototypes.
2920         * config/arm/vfp.md: Enable patterns for Thumb-2.
2921         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
2922         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2923         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2924         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2925         * config/arm/libunwind.S: Add Thumb-2 code.
2926         * config/arm/constraints.md: Update include Thumb-2.
2927         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2928         * config/arm/ieee754-df.S: Ditto.
2929         * config/arm/bpabi.S: Ditto.
2930         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2931         * config/arm/predicates.md (low_register_operand,
2932         low_reg_or_int_operand, thumb_16bit_operator): New.
2933         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2934         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2935         * config/arm/t-arm-elf: Add armv7 multilib.
2936         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2937         Include thumb2.md.
2938         (UNSPEC_STACK_ALIGN, ce_count): New.
2939         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2940         arm_zero_extendsidi2, arm_zero_extendqidi2): New
2941         insns/expanders.
2942         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2943         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2944         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2945         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2946         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2947         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2948         insns.
2949         * doc/extend.texi: Document ARMv7-M interrupt functions.
2950         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2951
2952 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
2953
2954         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2955         (struct _Unwind_Context): Rename args_size to flags, remove
2956         signal_frame field, add a new args_size field and version field.
2957         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2958         _Unwind_IsExtendedContext): New inline functions.
2959         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2960         Assume by_value array is only present if _Unwind_IsExtendedContext.
2961         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2962         _Unwind_IsSignalFrame.
2963         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2964         (uw_update_context_1): Use _Unwind_SetSignalFrame.
2965         (uw_init_context_1): Initialize context->flags to
2966         EXTENDED_CONTEXT_BIT.
2967         * config/rs6000/linux-unwind.h (frob_update_context): Use
2968         _Unwind_SetSignalFrame.
2969
2970 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
2971
2972         PR middle-end/30353
2973         * gimplify.c (gimplify_modify_expr_complex_part): Move below
2974         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
2975         the value.
2976
2977 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
2978
2979         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2980         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2981         RET, AT_SP): Remove.
2982         * config/i386/i386.md (*sse_prologue_save_insn): Use return
2983         instead of RET.
2984
2985         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2986         * cfgbuild.c (find_basic_blocks): Likewise.
2987         * cfgrtl.c (rtl_create_basic_block): Likewise.
2988         * function.c (temp_slots_at_level): Likewise.
2989         * reg-stack.c (stack_regs_mentioned): Likewise.
2990         * regclass.c (allocate_reg_info): Likewise.
2991         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2992         set_bb_for_stmt, move_block_to_fn): Likewise.
2993         * tree-complex.c (tree_lower_complex): Likewise.
2994         * vec.h (VEC_safe_grow_cleared): New.
2995
2996         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2997
2998 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
2999
3000         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3001         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3002         to loop_version.
3003         * cfgloopmanip.c (scale_loop_frequencies): Export.
3004         (loopify): Scale the frequencies by prescribed coefficients.
3005         (set_zero_probability): New function.
3006         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3007         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3008         and frequencies according to arguments.
3009         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3010         correctly.
3011         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3012         * modulo-sched.c (sms_schedule): Set probabilities for entering
3013         versioned loop correctly.
3014         * tree-vect-transform.c (vect_transform_loop): Ditto.
3015         * cfgloop.h (loopify, loop_version): Declaration changed.
3016         (scale_loop_frequencies): Declared.
3017
3018 2007-01-02  Jan Hubicka  <jh@suse.cz>
3019
3020         * cgraph.c: Include tree-flow.h
3021         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3022         early_local_passes.
3023         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3024         * tree-pass.h (pass_all_early_optimizations): Declare.
3025         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3026         early_local_passes.
3027         (cgraph_analyze_function): Do early_local_passes.
3028         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3029         Do not add referenced vars.
3030         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3031         SSA form.
3032         (gate_all_early_local_passes): New gate.
3033         (pass_early_local_passes): Use new gate.
3034         (execute_early_local_optimizations): New functions.
3035         (gate_all_early_optimizations): New gate.
3036         (pass_all_early_optimizations): New pass.
3037         (execute_free_datastructures): Free SSA only when initialized.
3038         (gate_init_datastructures): Init only when optimizing.
3039         (tree_lowering_passes): Do early local passes when called late.
3040         * tree-profile.c (do_tree_profiling): Don't profile functions added
3041         late.
3042         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3043         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3044         not active.
3045         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3046         during early_local_passes.
3047         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3048
3049
3050 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3051
3052         * Makefile.in: Update copyright year.
3053
3054 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3055
3056         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3057         in $(lang_checks) and check-consistency targets.
3058
3059 2007-01-02  Jan Hubicka  <jh@suse.cz>
3060
3061         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3062         Do not add referenced vars.
3063         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3064         are not active.
3065         * passes.c (init_optimization_passes): Put mudflap_2 after
3066         free_datastructures.
3067
3068 2007-01-02  Jan Hubicka  <jh@suse.cz>
3069
3070         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3071         Set NOTHROW flag on call statements proved to be nothrow.
3072         Update statement of local calls so new pure/const functions are
3073         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3074         goto receivers.
3075         (tree_rest_of_compilation): Register hooks and initialize bitmap
3076         early. Do not set after_inlining flag.
3077
3078 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3079
3080         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3081         HOST_BITS_PER_LONGLONG
3082
3083 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3084
3085         PR c/19977
3086         * c-typeck.c (store_init_value): Don't emit pedantic overflow
3087         warning for non-static initializers.
3088
3089 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
3090
3091         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3092         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3093         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3094         Always use set_unique_reg_note to add REG_EQUAL notes.
3095
3096 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3097
3098         Revert:
3099         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3100
3101         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3102         * cfgbuild.c (find_basic_blocks): Likewise.
3103         * cfgrtl.c (rtl_create_basic_block): Likewise.
3104         * function.c (temp_slots_at_level): Likewise.
3105         * reg-stack.c (stack_regs_mentioned): Likewise.
3106         * regclass.c (allocate_reg_info): Likewise.
3107         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3108         set_bb_for_stmt, move_block_to_fn): Likewise.
3109         * tree-complex.c (tree_lower_complex): Likewise.
3110         * vec.h (VEC_safe_grow_cleared): New.
3111
3112 2007-01-02  Ian Lance Taylor  <iant@google.com>
3113
3114         * c-common.c (c_common_truthvalue_conversion): When warning about
3115         using an assignment as a truth value, set TREE_NO_WARNING.
3116
3117 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3118
3119         PR middle-end/7651
3120         * c.opt (Wold-style-declaration): New.
3121         * doc/invoke.texi (C-only Warning Options): New.
3122         (Wold-style-declaration): Document it.
3123         (Wextra): Enabled by -Wextra.
3124         * c-opts.c (c_common_post_options): Enabled by -Wextra.
3125         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3126         -Wold-style-declaration.
3127
3128 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3129
3130         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3131         * cfgbuild.c (find_basic_blocks): Likewise.
3132         * cfgrtl.c (rtl_create_basic_block): Likewise.
3133         * function.c (temp_slots_at_level): Likewise.
3134         * reg-stack.c (stack_regs_mentioned): Likewise.
3135         * regclass.c (allocate_reg_info): Likewise.
3136         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3137         set_bb_for_stmt, move_block_to_fn): Likewise.
3138         * tree-complex.c (tree_lower_complex): Likewise.
3139         * vec.h (VEC_safe_grow_cleared): New.
3140
3141 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
3142
3143         * c-common.c (c_common_nodes_and_builtins): Since variants of
3144         void_type_node get built before it is given a name, we need to
3145         give those variants the name, too.
3146         (complete_array_type): We need to work with the canonical main
3147         type of the array, from which we will build the qualified version.
3148         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3149         * print-tree.c (print_node): Display canonical type information
3150         for each type.
3151         * stor-layout.c (layout_type): When we don't know the
3152         alignment of a type for which we're building an array, we end up
3153         guessing wrong, so make the type require structural equality.
3154         * tree.c (make_node_stat): When we build a new type, it is its
3155         own canonical type.
3156         (build_type_attribute_qual_variant): When building an attribute
3157         variant, its canonical type is the non-attribute variant. However,
3158         if the attributes are target-dependent and they differ, we need to
3159         use structural equality checks for this type.
3160         (build_qualified_type): A qualified type is not equivalent to its
3161         unqualified variant; set the canonical type appropriately.
3162         (build_distinct_type_copy): When building a distinct type from
3163         another type, the new type is its own canonical type.
3164         (build_variant_type_copy): When building a new type variant, we
3165         assume that it is equivalent to the original type.
3166         (build_pointer_type_for_mode): When building a pointer type, also
3167         build a canonical type pointer.
3168         (build_reference_type_for_mode): When building a reference type,
3169         also build a canonical type reference.
3170         (build_index_type): When we can't hash an index type (e.g.,
3171         because its maximum value is negative), the index type requires
3172         structural equality tests.
3173         (build_array_type): Build the canonical form of an array type.
3174         (build_function_type): Function types require structural equality,
3175         because they contain default arguments, attributes, etc.
3176         (build_method_type_directly): Ditto for method types.
3177         (build_offset_type): Build the canonical offset type.
3178         (build_complex_type): Build the canonical vector type.
3179         (make_vector_type): Build the canonical vector type.
3180         * tree.h (TYPE_CANONICAL): New.
3181         (TYPE_STRUCTURAL_EQUALITY_P): New.
3182         (SET_TYPE_STRUCTURAL_EQUALITY): New.
3183         (struct tree_type): Added "canonical" field.
3184         * params.h (VERIFY_CANONICAL_TYPES): New.
3185         * doc/c-tree.texi (TYPE_CANONICAL): Document.
3186         (TYPE_STRUCTURAL_EQUALITY_P): Document.
3187         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3188         * doc/invoke.texi (verify-canonical-types): Document --param
3189         parameter for verifying canonical types.
3190
3191 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3192
3193         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3194         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3195         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3196         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3197         tm_file.
3198         * config/rs6000/e500.h: New.
3199         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3200         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3201         Remove.
3202         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3203         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3204         TARGET_E500_DOUBLE): Remove.
3205         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3206         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3207         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3208         * config/rs6000/rs6000.c (rs6000_override_options): Use
3209         CHECK_E500_OPTIONS.
3210
3211 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3212
3213         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3214         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3215         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3216         rs6000_emit_cmove): Don't check TARGET_E500.
3217         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3218         sordered): Don't check TARGET_E500.
3219
3220 2007-01-01  Eric Christopher  <echristo@apple.com>
3221
3222         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3223         HARD_REGISTER_NUM_P.
3224
3225 2007-01-01  Roger Sayle  <roger@eyesopen.com>
3226
3227         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3228         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3229         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3230         eq/ne Y".
3231
3232 2007-01-01  Mike Stump  <mrs@apple.com>
3233
3234         * configure.ac: Remove support for building with Apple's gcc-3.1.
3235
3236 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3237
3238         PR middle-end/30311
3239         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3240         REG is a hard register.  Do not modify REG before calling
3241         subreg_nregs.
3242         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3243         multiple of the size of YMODE for certain lowpart cases.
3244
3245 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3246
3247         PR middle-end/30253
3248         * gimplify (voidify_wrapper_expr): Update for
3249         GIMPLE_MODIFY_STMT.
3250
3251 2007-01-01  Andreas Schwab  <schwab@suse.de>
3252
3253         PR target/29166
3254         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3255         save of BR0 in extra_spill_size instead of spill_size.
3256         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3257         area.
3258         (ia64_expand_epilogue): Restore BR0 from its new location.
3259
3260 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3261
3262         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3263         Use a temporary variable if the left hand side is not a gimple
3264         register.
3265
3266 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3267
3268         * gimplify.c (gimplify_return_expr): Make the temporary variable
3269         for the return expression, a gimple register variable.
3270
3271 2007-01-01  Jan Hubicka  <jh@suse.cz>
3272
3273         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3274         INSN_LIST.
3275
3276 2007-01-01  Mike Stump  <mrs@apple.com>
3277
3278         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3279         .literal16.
3280         * config/darwin.c (machopic_select_rtx_section): Use
3281         HAVE_GAS_LITERAL16.
3282         (darwin_mergeable_constant_section): Likewise.
3283         * configure: Regenerate.
3284         * config.in: Regenerate.
3285
3286 2007-01-01  Jan Hubicka  <jh@suse.cz>
3287             Andrew Pinski  <pinskia@gmail.com>
3288
3289         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3290         before starting IPA passes.