OSDN Git Service

f389a62056a3acdff849e0909e731a3342642a10
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
2
3         * tree.h (struct tree_memory_tag): Add aliases member.
4         (MTAG_ALIASES): New macro.
5         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
6         (add_may_alias): Remove pointer-set. Update for may_aliases being
7         a bitmap. 
8         (mark_aliases_call_clobbered): Update for may_aliases being a
9         bitmap.
10         (compute_tag_properties): Ditto.
11         (create_partition_for): Ditto.
12         (compute_memory_partitions): Ditto.
13         (dump_may_aliases_for): Ditto.
14         (is_aliased_with): Ditto.
15         (add_may_alias_for_new_tag): Ditto.
16         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
17         (compute_is_aliased): New function.
18         (compute_may_aliases): Call compute_is_aliased).
19         (init_alias_info): Initialize alias_bitmap_obstack.
20         (union_alias_set_into): New function.
21         (compute_flow_sensitive_aliasing): Use union_aliases_into.
22         (have_common_aliases_p): Rewrite to take two bitmaps and use
23         intersection.
24         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
25         Update for bitmaps.
26         (finalize_ref_all_pointers): Update for add_may_alias changes.
27         (new_type_alias): Ditto.
28         * tree-flow-inline.h (may_aliases): Return a bitmap.
29         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
30         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
31         may_aliases being a bitmap.
32         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
33         may_aliases now returns a bitmap.
34         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
35         being a bitmap.
36         * tree-ssa-operands.c (add_virtual_operand): Update for
37         may_aliases being a bitmap.
38         
39 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
40
41         PR tree-optimization/30630
42         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
43         variable as changed.
44
45 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
46
47         PR c++/28266
48         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
49         expanded only once even if an error occurs.
50
51 2007-01-29  Ben Elliston  <bje@au.ibm.com>
52
53         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
54         (__gcov_average_profiler): Likewise.
55         (__gcov_ior_profiler): Likewise.
56         (__gcov_merge_ior): Likewise.
57
58 2007-01-28  Jan Hubicka  <jh@suse.cz>
59
60         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
61         * value-prof.c (stringop_block_profile): Likewise.
62
63 2007-01-28  Jan Hubicka  <jh@suse.cz>
64
65         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
66         variant handling histograms; add wrapper.
67         (clear_storage_via_libcall): Export.
68         (emit_block_move_hints): Break out from ...; add histograms.
69         (emit_block_move): ... this one.
70         (clear_storage_hints): Break out from ...; add histograms.
71         (clear_storage): ... this one.
72         (set_storage_via_memset): Handle histogram.
73         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
74         (clear_storage_hints, clear_storage_via_libcall): Declare.
75         (set_storage_via_setmem): Update prototype.
76         * doc/md.texi (movmem, setmem): Document new arguments.
77
78         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
79         new histograms.
80         (stringop_block_profile): New global function.
81         (tree_stringops_values_to_profile): Profile block size and alignment.
82         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
83         HIST_TYPE_IOR.
84         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
85         (stringop_block_profile): Declare.
86         * builtins.c: Include value-prof.h.
87         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
88         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
89         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
90         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
91         * profile.c (instrument_values): Add new counters.
92         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
93         calls.
94         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
95         (tree_init_edge_profiler): Build new profilers.
96         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
97         (pass_tree_profile): Add dump.
98         (tree_profile_hooks): Update.
99         * Makefile.in (LIBGCOV): Add new constants.
100         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
101         __gcov_ior_profiler): New.
102         * i386.md (movmem/setmem expanders): Add new optional arguments.
103
104 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
105
106         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
107
108 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
109
110         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
111         (tree_unswitch_single_loop): Pass eni_size_weights to
112         tree_num_loop_insns.
113         * tree-ssa-loop-manip.c: Include tree-inline.h.
114         (can_unroll_loop_p): Pass eni_size_weights to
115         tree_num_loop_insns.
116         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
117         Pass eni_size_weights to estimate_num_insns.
118         * tree.h (init_inline_once): Export.
119         * toplev.c (backend_init): Call init_inline_once.
120         * cgraphunit.c (cgraph_process_new_functions,
121         cgraph_analyze_function): Pass eni_inlining_weights to
122         estimate_num_insns.
123         * ipa-inline.c (compute_inline_parameters): Ditto.
124         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
125         to estimate_num_insns.
126         (try_unroll_loop_completely): Pass eni_size_weights to
127         tree_num_loop_insns.
128         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
129         ot estimate_num_insns.
130         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
131         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
132         * tree-inline.c (eni_inlining_weights, eni_size_weights,
133         eni_time_weights): New variables.
134         (init_inline_once): Initialize them.
135         (struct eni_data): Mew.
136         (estimate_num_insns_1, estimate_num_insns): Use weights.
137         * tree-inline.h (struct eni_weights_d): New.
138         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
139         (estimate_num_insns): Declaration changed.
140         * cfgloop.h (tree_num_loop_insns): Declaration changed.
141         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
142         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
143
144 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
145
146         * tree-data-ref.c (conflict_fn): Assert that the number of affine
147         relations in the conflict function is valid.
148
149 2007-01-27  Ian Lance Taylor  <iant@google.com>
150
151         * common.opt: Add fstrict-overflow.
152         * opts.c (decode_options): Set flag_strict_overflow if -O2.
153         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
154         (TYPE_OVERFLOW_UNDEFINED): Define.
155         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
156         Replace all uses.
157         * tree.h (TYPE_TRAP_SIGNED): Don't define.
158         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
159         (fold_negate_expr): Likewise.
160         (make_range): Likewise.
161         (extract_muldiv_1): Likewise.
162         (maybe_canonicalize_comparison): Likewise.
163         (fold_comparison): Likewise.
164         (fold_binary): Likewise.
165         (tree_expr_nonnegative_p): Likewise.
166         (tree_expr_nonzero_p): Likewise.
167         * tree-vrp.c (compare_values): Likewise.
168         (extract_range_from_binary_expr): Likewise.
169         (extract_range_from_unary_expr): Likewise.
170         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
171         Likewise.
172         (nowrap_type_p): Likewise.
173         * tree-scalar-evolution.c (simple_iv): Likewise.
174         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
175         (build_range_check): Likewise.
176         (extract_muldiv_1): Likewise.
177         (fold_comparison): Likewise.
178         * tree-vrp.c (vrp_int_const_binop): Likewise.
179         (extract_range_from_unary_expr): Likewise.
180         * convert.c (convert_to_integer): Likewise.
181         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
182         (fold_comparison): Likewise.
183         (fold_binary): Likewise.
184         * optabs.c (optab_for_tree_code): Likewise.
185         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
186         * simplify-rtx.c (simplify_const_relational_operation): Check
187         flag_strict_overflow and flag_trapv.
188         (simplify_const_relational_operation): Likewise.
189         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
190         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
191         -fstrict-overflow.
192
193 2007-01-27  Roger Sayle  <roger@eyesopen.com>
194
195         * tree.c (tree_fold_gcd): Delete.
196         * tree.h (tree_fold_gcd): Remove prototype.
197         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
198         test whether one constant integer is a multiple of another.  Instead
199         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
200         * fold-const.c (multiple_of_p):  We've determined both TOP and
201         BOTTOM are integer constants so we can call int_const_binop directly
202         instead of the more generic const_binop.
203
204 2007-01-27  Roger Sayle  <roger@eyesopen.com>
205
206         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
207         1*X check that the constant hasn't overflowed, to preserve the
208         TREE_OVERFLOW bit.
209         (round_up): Provide an efficient implementation when rouding-up an
210         INTEGER_CST to a power-of-two.
211
212 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
213
214         * doc/sourcebuild.texi: Add comma for clarity.
215         * doc/extend.texi: Fix some typos.
216         * doc/passes.texi: Likewise.
217         * doc/cppinternals.texi: Likewise.
218         * doc/c-tree.texi: Likewise.
219         * doc/tree-ssa.texi: Likewise.
220         * doc/install.texi: Likewise.
221
222 2007-01-27  Jan Hubicka  <jh@suse.cz>
223
224         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
225         (pass_sra): Do not require alias information.
226         * passes.c (init_optimization_passes): Add SRA
227
228 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
229
230         * tracer.c (rest_of_handle_tracer): We already cleaned
231         up the CFG in tracer() so don't do it here again.
232         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
233         cleanup_cfg here, either.  And don't call renumber_insns.
234
235         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
236         and BB_HEAD are in the insn stream here.  Instead make sure
237         that BB_INSN is valid on all insns.  Also, do check here that
238         there are no pending branch predictions...
239         (rtl_verify_flow_info): ...instead of doing it here.  Checks
240         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
241         here.
242
243 2007-01-26  Roger Sayle  <roger@eyesopen.com>
244
245         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
246         function to simplify/factorize operand order canonicalization.
247         (ix86_fixup_binary_operands): Reorganize using the above function.
248         (ix86_binary_operator_ok): Likewise.
249
250 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
251
252         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
253         file scope from optimize_attrs.
254         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
255         up search.
256         (optimize_attrs): Clear insn_code_values after freeing it.
257
258 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
259
260         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
261         a parts.base assignment.
262
263 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
264
265         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
266         (compute_subscript_distance, initialize_data_dependence_relation,
267         finalize_ddr_dependent, analyze_ziv_subscript,
268         analyze_siv_subscript_cst_affine,
269         compute_overlap_steps_for_affine_univar,
270         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
271         analyze_siv_subscript, analyze_miv_subscript,
272         analyze_overlapping_iterations, subscript_dependence_tester_1,
273         compute_self_dependence, free_dependence_relation): Work
274         with affine_fn instead of chrecs.
275         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
276         common_affine_function, affine_function_base,
277         affine_function_constant_p, affine_fn_op, affine_fn_plus,
278         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
279         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
280         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
281         (all_chrecs_equal_p): Removed.
282         * tree-data-ref.h (affine_fn, conflict_function): New types.
283         (struct subscript): Change type of conflicting_iterations_in_a
284         and conflicting_iterations_in_b.
285
286 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
287
288         PR other/30182
289         * config/pa/pa.h (TARGET_HPUX_11): New.
290         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
291         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
292
293 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
294         Richard Guenther  <rguenther@suse.de>
295
296         * tree-ssa-structalias.c (solve_graph): Handle case
297         we merged the variable to another.
298
299 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
300
301         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
302         functions.
303
304         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
305         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
306
307 2007-01-25  DJ Delorie  <dj@redhat.com>
308
309         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
310         allow changes to modes which don't fit in those registers.
311
312         * reload1.c (choose_reload_regs): Check for invalid subregs before
313         computing their locations, not after.
314
315 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
316
317         PR 25127
318         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
319         save Altivec registers in an eh_return function.
320         (compute_vrsave_mask): Likewise.
321         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
322         for saving Altivec registers.
323         (rs6000_emit_prologue): Don't allocate stack twice in
324         eh_return function.  Correct expected value of altivec_save_offset
325         when using save_world.  Describe save of R0 to stack when using
326         save_world.  Describe stack pointer adjustment when using
327         save_world.  Remove duplicated eh_return parameter register saving.
328         Update sp_offset variable after save_world.
329         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
330         darwin-world.asm.
331         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
332         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
333         implies Altivec.
334
335 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
336
337         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
338
339 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
340
341         PR other/30182
342         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
343         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
344
345 2007-01-25  Richard Guenther  <rguenther@suse.de>
346
347         * doc/invoke.texi (-Wcoverage-mismatch): Document.
348         * common.opt (-Wcoverage-mismatch): New warning option.
349         * coverage.c (get_coverage_counts): Ignore coverage mismatch
350         if -Wcoverage-mismatch is given.
351
352 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
353
354         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA. 
355         (ipcp_driver): Change to static definition. 
356         Add dumping of the ifunctions.
357         (constant_val_insert): Remove unused parameter. Support for SSA.
358         (ipcp_propagate_const): Support for SSA.
359         (ipcp_profile_bb_print): Print only analyzed nodes.
360         (ipcp_replace_map_create): Remove support for Fortran constant 
361         for now.
362         * ipa-prop.c (ipa_method_modify_stmt, 
363         ipa_callsite_compute_param): Support for SSA.
364         * ipa-prop.h (ipcp_driver): Remove declaration.
365         (IS_VALID_TREE_MAP_INDEX): Add define.
366
367 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
368
369         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
370
371 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
372
373         PR middle-end/30447
374         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
375         constant argument to cabs and do it without checking for
376         -funsafe-math-optimizations.
377
378 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
379
380         * c-common.h (RID_FIRST_CXX0X): New.
381         (RID_LAST_CXX0X): New.
382         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
383         by -Wall.
384         * c.opt (Wc++0x-compat): New.
385         * doc/invoke.texi (-Wc++0x-compat): Document.
386
387 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
388
389         * builtins.c (fold_builtin_carg): New.
390         (fold_builtin_1): Use it.
391
392 2007-01-24  Jan Hubicka  <jh@suse.cz>
393
394         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
395
396         * ipa-inline.c (initial_insns, max_insns): Delete.
397         (compute_max_insns): New function.
398         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
399         of insns as base for code growth.
400         (cgraph_decide_inlining): Make initial_insns local; do not compute
401         max_insns.
402         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
403         * doc/invoke.texi (inline-unit-growth): Update docs.
404
405 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
406
407         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
408         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
409         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
410         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
411         for CPUs that have PTA_CX16 set.
412
413         PR middle-end/27416
414         * gimplify.c (omp_check_private): New function.
415         (gimplify_scan_omp_clauses): Use it for
416         firstprivate/lastprivate/reduction.
417
418         PR middle-end/30494
419         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
420         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
421
422         PR middle-end/30421
423         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
424         If lower_lastprivate_clauses emits some statements, append them
425         to dlist rather than body_p and to body_p append an initializer.
426         (lower_omp_for): Adjust caller.
427
428 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
429
430         * target.h (globalize_decl_name): New.
431         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
432         * output.h (default_globalize_decl_name): New.
433         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
434         globalize_label.
435         (globalize_decl): Ditto.
436         (default_globalize_decl_name): New.
437         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
438         (ia64_handle_version_id_attribute): New.
439         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
440         (ia64_asm_output_external): Use globalize_decl_name instead
441         of globalize_label.
442         * doc/extend.texi (version_id): New pragma.
443         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
444
445 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
446
447         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with 
448         _uleb128_t and _Unwind_SWord with _sleb128_t.
449         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
450         uw_frame_state_for, uw_update_context_1): Likewise.
451         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
452         * unwind-pe.h (read_uleb128, read_sleb128,
453         read_encoded_value_with_base): Likewise.
454         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
455
456 2007-01-24  Richard Guenther  <rguenther@suse.de>
457
458         * builtins.c (expand_builtin_cexpi): Get the fndecl
459         for cexp in the correct way.
460
461 2007-01-24  Jan Hubicka  <jh@suse.cz>
462
463         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
464
465 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
466
467         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
468         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
469         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
470         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
471         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
472         __SIZEOF_PTRDIFF_T__.
473         * doc/cpp.texi: Documentation for the new builtin macros added.
474
475 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
476
477         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
478         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
479         instead of operands[2] to avoid access past the end of array.
480
481 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
482
483         * reload1.c (emit_reload_insns): Pass the reload register
484         for a non-spill output reload through forget_old_reloads_1.
485
486 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
487
488         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
489         _SOFT_DOUBLE if doubles use software floating-point.
490         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
491         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
492         * config/rs6000/darwin-ldouble.c: Also compile functions for
493         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
494         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
495         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
496         commented-out long double override.
497         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
498         * config/rs6000/eabispe.h: Likewise.
499         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
500         long double for non-SPE.
501         (rs6000_handle_option): Likewise.
502         (invalid_e500_subreg): Disallow more subregs involding DImode,
503         DFmode, TImode or TFmode.
504         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
505         E500 double.
506         (legitimate_lo_sum_address_p): Also check for TFmode for E500
507         double.
508         (rs6000_legitimize_address): Also handle TFmode for E500 double.
509         (rs6000_legitimize_reload_address): Also handle TFmode for E500
510         double.
511         (rs6000_legitimate_address): Also check for TFmode for E500
512         double.
513         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
514         (spe_build_register_parallel): Handle TFmode and TCmode.
515         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
516         double.
517         (function_arg): Handle TFmode and TCmode for E500 double.
518         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
519         in general.
520         (print_operand): Handle TFmode and TImode for %y.
521         (rs6000_generate_compare): Handle TFmode comparisons for E500
522         double.
523         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
524         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
525         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
526         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
527         TFmode for E500 double.
528         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
529         (floatsidf2): Enable for E500 double.
530         (movtf_softfloat): Use rs6000_nonimmediate_operand.
531         (extenddftf2): Change to extenddftf2_fprs.
532         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
533         depending on TARGET_E500_DOUBLE.
534         (extendsftf2): Enable for E500 double.
535         (trunctfdf2): Enable for E500 double.
536         (trunctfsf2): Change to trunctfsf2_fprs.
537         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
538         depending on TARGET_E500_DOUBLE.
539         (floatsitf2): Enable for E500 double.
540         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
541         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
542         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
543         (negtf2): Change to negtf2_internal.
544         (negtf2): New expander.
545         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
546         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
547         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
548         (movdi_internal32): Use rs6000_nonimmediate_operand.
549         (unnamed splitter): Likewise.
550         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
551         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
552         (SPE64TF, DITI): New mode macros.
553         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
554         modes.
555         (frob_tf_ti): New.
556         (frob_<mode>_di_2): New.
557         (frob_tf_di_8_2): New.
558         (frob_di_df): Change to frob_di_<mode>; allow more modes.
559         (frob_ti_tf): New.
560         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
561         more modes.
562         (frob_ti_<mode>_8_2): New.
563         (frob_ti_tf_2): New.
564         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
565         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
566         TFmode.
567         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
568         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
569         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
570         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
571         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
572         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
573         tsttflt_gp): New.
574
575 2007-01-23  Ian Lance Taylor  <iant@google.com>
576
577         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
578         insn-*.o first.
579         (OBJS-archive): Reformat, alphabetize.
580         (OBJS): Change out_object_file to OBJS-md.
581
582 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
583
584         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
585         JUMP_P and CALL_P predicates where applicable.
586         * config/i386/i386.c: Ditto.
587         * config/i386/i386.md: Ditto.
588         * config/i386/mmx.md: Ditto.
589         * config/i386/predicates.md: Ditto.
590
591 2007-01-22  Andreas Schwab  <schwab@suse.de>
592
593         * config/m68k/m68k.h: Fix comment.
594
595 2007-01-22  Jan Hubicka  <jh@suse.cz>
596
597         * passes.c (init_optimization_passes): Do not rerun
598         pass_early_warn_uninitialized.
599
600 2007-01-22  Richard Guenther  <rguenther@suse.de>
601
602         PR tree-optimization/30038
603         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
604         function.
605         (execute_cse_sincos_1): Likewise.
606         (execute_cse_sincos): Likewise.
607         (gate_cse_sincos): Likewise.
608         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
609         the cexpi() canonicalization of sincos().
610         * tree-pass.h (pass_cse_sincos): Declare.
611         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
612
613 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
614
615         PR rtl-optimization/29329
616         * combine.c (replaced_rhs_insn): Rename to i2mod.
617         (replaced_rhs_value): Rename to i2mod_new_rhs.
618         (i2mod_old_rhs): New global variable.
619         (combine_instructions): Adjust for above change.  Save a copy of
620         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
621         note are substituted in the second instruction.
622         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
623         ditch the note if it pertains to the second eliminated register
624         and this register is mentioned in i2mod_old_rhs.
625
626         Revert:
627         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
628
629         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
630         the insn to which the note was originally attached.
631
632 2007-01-21  Jan Hubicka  <jh@suse.cz>
633
634         * ipa-inline.c (inlining_mode): Comment, move up.
635         (cgraph_decide_inlining_incrementally): Do not perform inlining
636         itself; fix handling of flattening of self recursive functions.
637         (cgraph_find_cycles): Remove.
638         (cgraph_flatten_node): Remove.
639         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
640         (try_inline): New function.
641         (cgraph_early_inlining): Update call of
642         cgraph_decide_inlining_incrementally.  Apply inlining here.
643         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
644
645 2007-01-21  Dirk Mueller  <dmueller@suse.de>
646
647         PR bootstrap/30511
648         * tree-vrp.c (check_array_bounds): do not warn
649         about ADDR_EXPR's of ARRAY_REF's which are immediately
650         used in binary expressions.
651
652 2007-01-21  Ira Rosen  <irar@il.ibm.com>
653
654         * tree-vectorizer.h (struct _stmt_vec_info): Add new field 
655         read_write_dep and macros for its access.
656         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
657         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove 
658         argument, call vect_check_interleaving for every independent pair of
659         data-refs. Mark loads that access the same memory location as a store
660         in the loop.
661         (vect_check_dependences): Remove.
662         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences 
663         call, fix the call to vect_analyze_data_ref_dependence.
664         (vect_analyze_data_ref_access): For statements that access the same 
665         data-ref, check that they are not stores; for loads, check that there
666         is no store that access the same location.
667
668 2007-01-20  Roger Sayle  <roger@eyesopen.com>
669             Joseph Myers  <joseph@codesourcery.com>
670
671         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
672         (-fdump-rtl-stack): Likewise.
673         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
674         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
675         (-mpt-fixed): Use @minus{} for minus sign.
676         (Using Precompiled Headers): Fix under/overfull hbox.
677
678 2007-01-20  Jan Hubicka  <jh@suse.cz>
679
680         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
681         place within annotation.
682
683 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
684
685         * pa.c (output_move_double): Change array size of xoperands to 4.
686
687 2007-01-20  Jan Hubicka  <jh@suse.cz>
688
689         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
690         as temporary.
691
692 2007-01-19  Ian Lance Taylor  <iant@google.com>
693
694         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
695         constant to unsigned HOST_WIDE_INT before negating.
696
697 2007-01-19  Ian Lance Taylor  <iant@google.com>
698
699         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
700         to unsigned.
701         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
702         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
703         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
704         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
705         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
706         (realloc_vdef, realloc_vuse): Change second parameter to
707         unsigned.
708         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
709         fields to unsigned.
710         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
711         unsigned. Change x and lim locals to unsigned.
712         (realloc_vdef, realloc_vuse): Change num_elem parameter to
713         unsigned.
714         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
715         (copy_virtual_operands): Change i and n locals to unsigned.
716
717 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
718
719         PR middle-end/29335
720         * builtins.c (fold_builtin_1): Handle builtin fdim.
721
722 2007-01-20  Jan Hubicka  <jh@suse.cz>
723
724         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
725         * tree-ssa-structalias.c (compute_points_to_sets): Do call
726         init_alias_heapvars.
727         (init_alias_heapvars): Initialize only when not already initialized.
728         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
729
730 2007-01-19  Roger Sayle  <roger@eyesopen.com>
731
732         * common.opt (fsigned-zeros): New command line option.
733         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
734         of flag_unsafe_math_optimizations.
735         * opts.c (set_fast_math_flags): The -ffast-math command line option
736         implies -fno-signed-zeros.
737         (fast_math_flags_set_p): Likewise.
738
739         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
740         the documentation of -ffast-math appropriately.  Wrap long lines.
741
742 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
743
744         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
745         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
746
747 2007-01-19  Tomas Bily  <tbily@suse.cz>
748
749         * cgraphunit.c (cgraph_finalize_function): Updating of pid
750         * tree-profile.c:
751         (tree_init_ic_make_global_vars): New function
752         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
753         (tree_gen_ic_profiler): New function
754         (tree_gen_ic_func_profiler): New function
755         (tree_profiling): Added calling of tree_gen_ic_func_profiler
756         (tree_profile_hooks): Added hook for indirec/virtual calls
757         * value-prof.c (tree_find_values_to_profile): New case for
758         indirect calls
759         (tree_values_to_profile): Call for determining indirect/virtual
760         counters
761         (tree_indirect_call_to_profile): New function
762         (tree_ic_transform): New function
763         (tree_ic): New function
764         (find_func_by_pid): New function
765         (init_pid_map): New function
766         (tree_value_profile_transformations): Added check for
767         indirect/virtual call transformation
768         * value-prof.h (enum hist_type): New counter type for
769         indirect/virtual calls
770         (profile_hooks): Added new hook for profiling indirect/virtual
771         calls
772         * profile.c (instrument_values): New case for indirect/virtual
773         call added
774         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
775         (GCOV_COUNTER_V_INDIR): New counter type
776         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
777         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
778         call added
779         * cgraph.c: Definition of cgraph_max_pid
780         (cgraph_create_node): Default init of pid attribute
781         * cgraph.h: Declaration of cgraph_max_pid
782         (struct cgraph_node): Added pid attribute
783         * libgcov.c (__gcov_indirect_call_profiler): New function
784         (__gcov_one_value_profiler_body): New function
785         (__gcov_one_value_profiler): Body was moved to
786         __gcov_one_value_profiler_body and calls it
787
788 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
789         
790         * doc/gty.texi (Options): Document the mark_hook option to GTY.
791         * gengtype.c (write_types_data, write_func_for_structure,
792         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
793         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
794         (walk_type, write_func_for_structure): Generate the mark_hook if
795         needed.
796
797 2007-01-19  Jan Hubicka  <jh@suse.cz>
798
799         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
800         'early' argument take inlining mode argument specifying whether to
801         inline for size/speeed or all functions; add support for flattening;
802         improve dumpting.
803         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
804
805 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
806
807         PR c++/17947
808         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
809         the name of the declared identifier.
810                 
811 2007-01-19  Dirk Mueller  <dmueller@suse.de>
812
813         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
814         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
815         duplicate evaluation.
816
817 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
818
819         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
820         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
821         to XFMode, use acosxf2 and truncate result to requested mode.
822         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
823         (asin<mode>2): Similarly, with asin expanders.
824         (*fscalexf4_i387): Rename from *fscalexf4.
825         (expNcorexf3): New expander.
826         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
827         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
828         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
829         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
830         patterns for SSE math.
831         (exp10<mode>2): Similarly, with exp10 expanders.
832         (exp2<mode>2): Similarly, with exp2 expanders.
833         (expm1<mode>2): Similarly, with expm1 expanders.
834         (ldexp<mode>3): Similarly, with ldexp expanders.
835         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
836         Use gen_truncxf<mode>2_i387_noop to truncate result.
837
838 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
839
840         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
841         for ColdFire targets.
842
843 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
844             Richard Sandiford  <richard@codesourcery.com>
845
846         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
847         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
848         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
849         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
850         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
851         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
852
853 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
854
855         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
856         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
857         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
858         (*adddi_dilshr32_cf): New pattern.
859         (adddi3, subdi3): Remove first alternatives.
860
861 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
862
863         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
864         implemented using move.l, do not use its cc result for floating-point
865         comparisons.
866
867 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
868
869         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
870         from accepting 's' constraints if flag_pic.
871
872 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
873
874         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
875         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
876         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
877         condition from TARGET_68881 to TARGET_HARD_FLOAT.
878
879 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
880
881         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
882         ColdFire alternatives.
883         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
884         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
885         architectures that support ff1 instruction.
886
887 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
888             Julian Brown  <julian@codesourcery.com>
889
890         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
891         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
892         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
893         Add commentary.
894
895 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
896
897         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
898         and return a bool.
899         (output_move_const_into_data_reg, output_move_simode_const): Delete.
900         * config/m68k/m68k.c (const_method, const_int_cost): Take a
901         HOST_WIDE_INT instead of an rtx.
902         (m68k_rtx_costs): Update call accordingly.
903         (output_move_const_into_data_reg): Likewise.  Fix formatting.
904         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
905         Return a bool.
906         (output_move_simode_const): Update calls after above changes.
907         Rework to use automatic variables and predicates like MEM_P.
908         * config/m68k/m68k.md (pushexthisi_const): Update call to
909         valid_mov3q_const.
910
911 2007-01-19  Dirk Mueller  <dmueller@suse.de>
912
913         * tree-ssa-alias.c (perform_var_substitution): Fix typo
914         in dump_flags test.
915
916 2007-01-19  Richard Guenther  <rguenther@suse.de>
917
918         * builtins.c (expand_builtin_cexpi): Fall back to expanding
919         via cexp in case sincos is not available.
920
921 2007-01-19  Richard Guenther  <rguenther@suse.de>
922
923         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
924         * defaults.h (TARGET_HAS_SINCOS): Default to off.
925         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
926         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
927         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
928         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
929         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
930         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
931
932 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
933
934         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
935         New insn patterns.
936         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
937         (atan2xf3): Directly generate RTL pattern.
938         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
939         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
940         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
941         disable patterns for SSE math.
942         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
943         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
944         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
945         disable patterns for SSE math.
946
947 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
948
949         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
950
951 2007-01-18  Roger Sayle  <roger@eyesopen.com>
952
953         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
954         VIEW_CONVERT_EXPR to the same type as it's operand.
955
956 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
957
958         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
959         configured for long double 128.
960
961 2007-01-18  Mike Stump  <mrs@apple.com>
962
963         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
964         error.
965
966 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
967
968         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
969         type correct.
970
971 2007-01-18  Jan Hubicka  <jh@suse.cz>
972
973         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
974         greater than NUM_VOP_FREE_BUCKETS.
975
976 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
977
978         * tree-ssa-structalias.c: Update comments.
979         (ptabitmap_obstack): Removed.
980         (pta_obstack): New.
981         (oldpta_obstack): Ditto.
982         (stats): Add a few members.
983         (struct variable_info): Remove node, complex, address_taken, and
984         indirect_target members. Add oldsolution member.
985         (new_var_info): Do not initialize removed members.
986         (constraint_expr_type): Remove INCLUDES.
987         (constraint_graph): Add size, implicit_preds, rep,
988         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
989         (FIRST_REF_NODE): New macro.
990         (LAST_REF_NODE): Ditto.
991         (FIRST_ADDR_NODE): Ditto.
992         (find): New function.
993         (unite): Ditto.
994         (dump_constraint): Do not handle INCLUDES.
995         (insert_into_complex): Do not insert duplicate constraints.
996         (condense_varmap_nodes): Renamed and rewritten into ...
997         (merge_node_constraints): This. Also fix bug in handling of
998         offseted copy constraints.
999         (clear_edges_for_node): No longer need to deal with preds at all,
1000         or removing associated preds/succs. 
1001         (merge_graph_nodes): Deal with indirect_cycles.
1002         Don't deal with predecessors.
1003         (add_implicit_graph_edge): New function.
1004         (add_pred_graph_edge): Ditto.
1005         (add_graph_edge): Don't deal with predecessors.
1006         (build_constraint_graph): Removed.
1007         (build_pred_graph): New function.
1008         (build_succ_graph): Ditto.
1009         (struct scc_info): Removed in_component. Added roots, dfs, and
1010         node_mapping. Remove visited_index, unification_queue.
1011         (scc_visit): Deal with union-find we do now.
1012         Deal with cycles with REF nodes.
1013         (collapse_nodes): Renamed and rewritten to ...
1014         (unify_nodes): This.
1015         (process_unification_queue): Removed.
1016         (topo_visit): Cleanup
1017         (do_da_constraint): Use find.
1018         (do_sd_constraint): Ditto.
1019         (do_ds_constraint): Ditto.
1020         (do_complex_constraint): Ditto.
1021         (init_scc_info): Update for removed and added members.
1022         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1023         (find_indirect_cycles): This.
1024         (equivalence_class): New variable.
1025         (label_visit): New function.
1026         (perform_variable_substitution): Rewritten.
1027         (free_var_substitution_info): New function.
1028         (find_equivalent_node): Ditto.
1029         (move_complex_constraints): Ditto.
1030         (eliminate_indirect_cycles): Ditto.
1031         (solve_graph): Only propagate changed bits.
1032         Use indirect cycle elimination.
1033         Use find.
1034         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1035         (tree_id_eq): Renamed to ...
1036         (tree_vi_eq): This. Update for member change
1037         (insert_id_for_tree): Renamed and rewritten to ...
1038         (insert_vi_for_tree): This.
1039         (lookup_id_for_tree): Renamed and rewritten to ...
1040         (lookup_vi_for_tree): This.
1041         (get_id_for_tree): Renamed and rewritten to ...
1042         (get_vi_for_tree): Ditto.
1043         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1044         (process_constraint): Don't handle INCLUDES.
1045         Remove special ADDRESSOF case.
1046         (find_func_aliases): Rewrite to use vi functions instead of id
1047         ones.
1048         (create_function_info_for): Ditto.
1049         (create_variable_info_for): Ditto.
1050         (intra_create_variable_infos): Ditto.
1051         (merge_smts_into): Ditto.
1052         (find_what_p_points_to): Ditto.
1053         (init_base_vars): Ditto.
1054         (init_alias_vars): Ditto.
1055         (remove_preds_and_fake_succs): New function.
1056         (dump_sa_points_to_info): Dump new stats.
1057         (dump_solution_for_var): Use find.
1058         (set_used_smts): Fix formatting.
1059         (compute_points_to_sets): Updated for new functions.
1060         (ipa_pta_execute): Ditto.
1061
1062 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1063             Richard Sandiford  <richard@codesourcery.com>
1064
1065         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1066         parallels.
1067         * calls.c (expand_call): If the return value is a PARALLEL,
1068         extract its first member.
1069         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1070         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1071         use FUNCTION_EXTRA_EPILOGUE.
1072         (m68k_function_value): Return a PARALLEL if the return value
1073         is of a pointer type.
1074         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1075         (FUNCTION_EXTRA_EPILOGUE): Remove.
1076         * config/m68k/m68k.md (D0_REG): New constant.
1077
1078 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1079
1080         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1081         output a NOP for empty epilogues.
1082
1083 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1084
1085         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1086         before function.  Extend register save check to include all
1087         registers, not just integer ones.
1088
1089 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1090
1091         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1092         (m68k_use_return_insn): ...this.
1093         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1094         * config/m68k/m68k.c (use_return_insn): Rename to...
1095         (m68k_use_return_insn): ...this.
1096         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1097         of USE_RETURN_INSN.
1098
1099 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1100
1101         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1102         nearest even, fix denormal rounding overflow.
1103
1104 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1105
1106         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1107         f%$smove and f%$move.
1108         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1109         (extendsfdf2_cf): Use fdmove instead of f%&move.
1110         (truncdfsf2_cf): Use fsmove instead of f%$smove.
1111         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1112
1113 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1114
1115         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1116         GPR<-GPR moves.
1117
1118 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1119
1120         * real.h (real_format): Add a canonical_nan_lsbs_set field.
1121         (coldfire_single_format): Declare.
1122         (coldfire_double_format): Likewise.
1123         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1124         of qnan_msb_set to determine the lower bits of a canonical
1125         NaN significand.
1126         (encode_ieee_double): Likewise.
1127         (encode_ieee_quad): Likewise.
1128         (ieee_single_format): Initialize canonical_nan_lsbs_set.
1129         (mips_single_format): Likewise.
1130         (ieee_double_format): Likewise.
1131         (mips_double_format): Likewise.
1132         (ieee_extended_motorola_format): Likewise.
1133         (ieee_extended_intel_96_format): Likewise.
1134         (ieee_extended_intel_128_format): Likewise.
1135         (ieee_extended_intel_96_round_53_format): Likewise.
1136         (ibm_extended_format): Likewise.
1137         (mips_extended_format): Likewise.
1138         (ieee_quad_format): Likewise.
1139         (mips_quad_format): Likewise.
1140         (vax_f_format): Likewise.
1141         (vax_d_format): Likewise.
1142         (vax_g_format): Likewise.
1143         (i370_single_format): Likewise.
1144         (i370_double_format): Likewise.
1145         (decimal_single_format): Likewise.
1146         (decimal_double_format): Likewise.
1147         (decimal_quad_format): Likewise.
1148         (c4x_single_format): Likewise.
1149         (c4x_extended_format): Likewise.
1150         (real_internal_format): Likewise.
1151         (coldfire_single_format): New real_format.
1152         (coldfire_double_format): Likewise.
1153         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1154         canonical_nan_lsbs_set.
1155         (pdp11_d_format): Likewise.
1156         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1157         if TARGET_COLDFIRE_CPU.
1158
1159 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1160
1161         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1162         (output_call, m68k_legitimize_call_address): Declare.
1163         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1164         parenthesees.  Add support for a 'W' constraint.
1165         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1166         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1167         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1168         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1169         variables.
1170         (override_options): Initialize them.  Do not set flag_no_function_cse
1171         for TARGET_ID_SHARED_LIBRARY.
1172         (m68k_output_pic_call): Delete.
1173         (m68k_legitimize_call_address): New function.
1174         (print_operand): Remove the %o prefix.  Handle the %p prefix.
1175         (output_call): New function.
1176         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
1177         target address from the GOT if symbolic jumps are not allowed.
1178         * config/m68k/m68k.md (call, general_operand): Do not set
1179         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
1180         Merge separate flag_pic and !flag_pic define_insns into...
1181         (*call, *call_value): ...these new patterns.  Match the address
1182         rather than the containing MEM and require it to be a call_operand.
1183         Use output_call to generate the asm template.
1184         * config/m68k/predicates.md (const_call_operand): New predicate.
1185         (call_operand): Likewise.
1186
1187 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1188
1189         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1190         (STACK_GROWS_DOWNWARD): Define to 1.
1191         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1192         comments, emphasizing that these values are only defaults.
1193         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1194         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1195
1196 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1197
1198         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1199
1200 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1201             Richard Sandiford  <richard@codesourcery.com>
1202
1203         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1204         (FP_REGNO_P): New macros.
1205         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1206         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1207         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1208         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1209
1210 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1211
1212         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1213         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1214         to $tmake_file.
1215         * config/m68k/t-floatlib: New file, extracting common code from...
1216         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1217         * config/m68k/t-uclinux: Here.
1218         * config/m68k/fpgnulib.c: Do not compile extendeed precision
1219         routines on ColdFire targets.
1220
1221 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1222
1223         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1224         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1225         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1226         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1227         explicitly set MOTOROLA to 1.
1228         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1229
1230 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1231             Nathan Sidwell  <nathan@codesourcery.com>
1232
1233         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1234         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1235         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1236         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1237         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1238         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1239         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1240         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1241         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1242         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1243         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1244         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1245         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1246         (DBX_REGISTER_NUMBER): Delete.
1247         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1248         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1249         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1250
1251 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1252
1253         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1254         * config/flat.h: New file.
1255         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1256         OBJECT_FORMAT_FLAT.
1257         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1258         indirect references for -msep-data or -mid-shared-library.
1259         Do not use PC-relative code addresses either.
1260         * config/m68k/m68k.c (override_options): Restrict -fPIC error
1261         to -mpcrel.
1262         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
1263         for shared libraries and crt1.o for executables.  Use crti.o and
1264         crtbegin.o.
1265         (ENDFILE_SPEC): Use crtend.o and crtn.o.
1266         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1267         Do not add -elf2flt or -shared-lib-id options here.
1268         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
1269         Pass -shared-lib-id if -mid-shared-library, taking the library
1270         identifier from -mshared-library-id if given, otherwise
1271         defaulting to 0.
1272         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1273         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1274         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1275         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1276         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1277         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1278         and crtend.o.
1279         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1280         jump sequence for ISA A and ISA A+.
1281         (PICJUMP): Likewise.
1282
1283 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1284
1285         * config.gcc (m68k-*-uclinux*): Base the port on the common
1286         and m68k GNU/Linux files rather than on the generic ELF ones.
1287         * config/m68k/uclinux.h (TARGET_VERSION): Override.
1288         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1289
1290 2007-01-18  Julian Brown  <julian@codesourcery.com>
1291             Richard Sandiford  <richard@codesourcery.com>
1292
1293         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1294         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1295         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1296         before redefining.
1297         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1298         80 unconditionally.
1299         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1300
1301 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1302
1303         * doc/install.texi: Document m68k-uclinuxoldabi.
1304         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1305         * config/m68k/uclinux-oldabi.h: New file, copied from
1306         config/m68k/uclinux.h.
1307
1308 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
1309
1310         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1311         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1312         (m32r_compute_frame_size): Use unsigned for regno.
1313         (m32r_reload_lr): Use gen_frame_mem.
1314         (pop): New.
1315         (m32r_output_function_epilogue): Don't output the function epilogue
1316         textually here.
1317         (m32r_expand_epilogue): New.
1318         (direct_return): Return false if the function has the interrupt
1319         attribute.
1320         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1321         * config/m32r/m32r.md (epilogue): New expander.
1322         (return_lr, return_rte): New insns.
1323         (return): Make it expander.
1324         (return_normal): New expander.
1325
1326 2007-01-18  Josh Conner  <jconner@apple.com>
1327
1328         PR target/30485
1329         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1330         support for UNLE, UNLT, UNGE, and UNGT.
1331
1332 2007-01-18  Jan Hubicka  <jh@suse.cz>
1333
1334         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1335         * tree-tailcall (add_virtual_phis): Likewise.
1336         (optimize_tail_call): Return TODOs.
1337         (execute_tail_calls): Return TODOs.
1338         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1339         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1340         changed.
1341         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1342         Cleanup cfg using TODO.
1343         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1344         * passes.c (execute_function_todo): When cleanup did something, remove
1345         unused locals.
1346         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1347         (make_edges): Don't cleanup_cfg.
1348
1349 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
1350
1351         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1352         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1353         operate on the top of stack.
1354         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1355         register handling of instructions that output to the second
1356         stack slot.
1357         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1358         (move_for_stack_reg): Special-case check for dead destination
1359         stack slot for constant load of 1.0 inside UNSPEC_TAN.
1360
1361         * config/i386/i386.md (UNSPEC_TAN): New constant.
1362         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1363         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1364         to correctly model move of constant 1.0 to top stack slot.
1365         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1366         (unnamed peephole2 pattern): Remove corresponding peephole2
1367         pattern that optimizes tan insn and loading of constant 1.0.
1368         (tanxf2): Use fptanxf4_i387.
1369         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1370         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
1371         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1372         disable patterns for SSE math.
1373         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1374         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1375
1376 2007-01-18  Dirk Mueller  <dmueller@suse.de>
1377             Richard Guenther <rguenther@suse.de>
1378
1379         PR diagnostic/8268
1380         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1381         * common.opt (Warray-bounds): Add new warning option.
1382         * c-opts.c (c_common_handle_option): Define -Warray-bounds
1383         if -Wall is given.
1384         * Makefile.in: make tree-vrp.o depend on toplev.h
1385         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1386         is enabled.
1387         (check_array_refs, check_array_bounds, check_array_ref): New.
1388
1389 2007-01-18  Jan Hubicka  <jh@suse.cz>
1390
1391         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1392         (execute_ssa_ccp): Return flags conditionally.
1393         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1394         changed.
1395         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1396
1397 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
1398
1399         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1400         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1401
1402         * Makefile.c (GTFILES): Add cfglayout.h.
1403         * gengtype.c (open_base_files): Likewise.
1404         * cfglayout.c (cfg_layout_function_footer,
1405         cfg_layout_function_header) Reindent to make gengtype happy.
1406         * cfglayout.h (cfg_layout_function_footer,
1407         cfg_layout_function_header): Add GTY(()) marker.
1408
1409         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1410
1411 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1412
1413         * genautomata.c (write_automata): Include xstrerror output in the
1414         error message if writing the DFA description file fails.
1415
1416 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1417
1418         * config/mips/mips-protos.h (mips_output_external): Make it
1419         return void.
1420         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1421         * config/mips/mips.c (irix_output_external_libcall): Likewise.
1422         (extern_list): Likewise.
1423         (extern_head): Likewise.
1424         (TARGET_ASM_FILE_END): Likewise.
1425         (mips_file_end): Likewise.
1426         (mips_output_external): Rewritten.
1427
1428 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1429
1430         * genpreds.c (write_insn_preds_c): Only write out the function
1431         body for regclass_for_constraint if we have register constraints.
1432
1433 2007-01-17  Tom Tromey  <tromey@redhat.com>
1434
1435         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1436         * doc/install.texi (Testing): Use sourceware.org.
1437         (Binaries): Likewise.
1438         (Specific): Likewise.
1439         * doc/contrib.texi (Contributors): Use sourceware.org.
1440
1441 2007-01-17  Anatoly Sokolov <aesok@post.ru>
1442
1443         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1444         (AVR_ENHANCED): Rename to ...
1445         (AVR_HAVE_MUL): ... new.
1446         (avr_enhanced_p): Rename to ...
1447         (avr_have_mul_p): ... new.
1448         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of 
1449         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1450         * config/avr/avr.c (avr_enhanced_p): Rename to ...
1451         (avr_have_mul_p): ... new.
1452         (base_arch_s): Rename 'enhanced' to 'have_mul'.
1453         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1454         'avr_enhanced_p' and 'enhanced'.
1455         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use 
1456         AVR_HAVE_MUL instead of AVR_ENHANCED.
1457         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1458         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3, 
1459         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL 
1460         instead of AVR_ENHANCED.
1461         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1462         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1463         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1464
1465 2007-01-17  Ian Lance Taylor  <iant@google.com>
1466
1467         * vec.h (VEC_reserve_exact): Define.
1468         (vec_gc_p_reserve_exact): Declare.
1469         (vec_gc_o_reserve_exact): Declare.
1470         (vec_heap_p_reserve_exact): Declare.
1471         (vec_heap_o_reserve_exact): Declare.
1472         (VEC_OP (T,A,reserve_exact)): New static inline function, three
1473         versions.
1474         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1475         negative parameter.
1476         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1477         (VEC_OP (T,A,copy)) [all versions]: Likewise.
1478         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1479         * vec.c (calculate_allocation): Add exact parameter.  Change all
1480         callers.
1481         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1482         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1483         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1484         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1485         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1486         (vec_heap_p_reserve_exact): New function.
1487         (vec_heap_o_reserve_exact): New function.
1488
1489 2007-01-17  Jan Hubicka  <jh@suse.cz>
1490
1491         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1492
1493 2007-01-17  Eric Christopher  <echristo@apple.com>
1494
1495         * config.gcc: Support core2 processor.
1496
1497 2007-01-16  Jan Hubicka  <jh@suse.cz>
1498
1499         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1500         the names, just unlink the chain so we don't crash on dangling pointers
1501         to dead SSA names.
1502
1503 2007-01-16  Jan Hubicka  <jh@suse.cz>
1504
1505         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1506         * tree-pass.h: Reorder to make IPA passes appear toegher.
1507         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1508         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1509         parameters, do not call early inliner.
1510         * ipa-inline.c: Update comments.  Include tree-flow.h
1511         (cgraph_decide_inlining): Do not compute inlining parameters.
1512         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1513         be called with function context set up.
1514         (pass_ipa_inline): Remove unreachable functions before pass.
1515         (cgraph_early_inlining): Simplify assuming to be called from the
1516         PM as local pass.
1517         (pass_early_inline): New pass.
1518         (cgraph_gate_ipa_early_inlining): New gate.
1519         (pass_ipa_early_inline): Turn into simple wrapper.
1520         (compute_inline_parameters): New function.
1521         (gate_inline_passes): New gate.
1522         (pass_inline_parameters): New pass.
1523         (apply_inline): Move here from tree-optimize.c
1524         (pass_apply_inline): New pass.
1525         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1526         transforming.
1527         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1528         doing them by hand.
1529         (tree_function_versioning): Do not allocate dummy struct function.
1530         * tree-inline.h (optimize_inline_calls): Update prototype.
1531         * tree-optimize.c (execute_fixup_cfg): Export.
1532         (pass_fixup_cfg): Remove
1533         (tree_rest_of_compilation): Do not apply inlines.
1534         * tree-flow.h (execute_fixup_cfg): Declare.
1535         * Makefile.in (gt-passes.c): New.
1536         * passes.c: Include gt-passes.h
1537         (init_optimization_passes): New passes.
1538         (nnodes, order): New static vars.
1539         (do_per_function_toporder): New function.
1540         (execute_one_pass): Dump current pass here.
1541         (execute_ipa_pass_list): Don't dump current pass here.
1542
1543 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
1544
1545         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1546         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1547
1548 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
1549
1550         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1551         (strong_alias): Define.
1552         (__gcc_qmul): Provide non-FMA for soft-float.
1553         (__gcc_qdiv): Same.
1554         (__gcc_qneg): New.
1555         (__gcc_qeq): New.
1556         (__gcc_qle): New.
1557         (__gcc_qge): New.
1558         (__gcc_qunord): New.
1559         (__gcc_stoq): New.
1560         (__gcc_dtoq): New.
1561         (__gcc_qtos): New.
1562         (__gcc_qtod): New.
1563         (__gcc_qtoi): New.
1564         (__gcc_qtou): New.
1565         (__gcc_itoq): New.
1566         (__gcc_utoq): New.
1567         (fmsub): New.
1568         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1569         soft-float functions.
1570         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1571         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1572         about long double soft float.
1573
1574 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
1575             Tehila Meyzels  <tehila@il.ibm.com>
1576
1577         * tree-vectorizer.h (is_pattern_stmt_p): New.
1578         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1579         formatting (tabs instead of spaces). Cleanup and clarify setting
1580         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1581         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1582         (vectorizable_type_demotion): Check that types are integral.
1583         (vectorizable_type_promotion): Likewise.
1584         (vectorizable_store): Fix typo.  Eliminate new-line at end of
1585         comments.
1586
1587 2007-01-16  Jan Hubicka  <jh@suse.cz>
1588
1589         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1590         cgraph edges too.
1591
1592 2007-01-15  Eric Christopher  <echristo@apple.com>
1593
1594         * ifcvt.c: Include vec.h, vecprim.h.
1595         (check_cond_move_block): New argument regs.
1596         Reorganize. Add registers used to regs.
1597         (cond_move_process_if_block): Use regs set above as
1598         loop bounds.
1599
1600 2007-01-15  Eric Christopher  <echristo@apple.com>
1601
1602         * config/darwin.h: Update copyright.
1603         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1604         (LINK_COMMAND_SPEC): Add -u.
1605         (LINK_SPEC): Fix umbrella for above.
1606
1607 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
1608
1609         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1610         glibc CVS.
1611
1612 2007-01-15  Tom Tromey  <tromey@redhat.com>
1613
1614         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1615         * doc/install.texi (Testing): Don't mention jacks.
1616         (Configuration): Document --enable-java-maintainer-mode.  Move
1617         --with-java-home to libgcj-specific section.  Document
1618         --with-ecj-jar.
1619         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1620
1621 2007-01-15  Jan Hubicka  <jh@suse.cz>
1622
1623         * tree-ssa-dce.c (DCE_TODOs): New.
1624         (propagate_necessity): Return if something changed.
1625         (eliminate_unnecessary_stmts): Likewise.
1626         (perform_tree_ssa_dce): Return TODO flags when needed.
1627         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1628
1629 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
1630
1631         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1632         (fyl2x_extend<mode>xf3_i387): New insn pattern.
1633         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1634         patterns using X87MODEF12 mode macro.  Extend operand 1
1635         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1636         (log10<mode>2): Ditto.
1637         (log2<mode>2): Ditto.
1638         (log1p<mode>2): Ditto.
1639         (logb<mode>2): Ditto.
1640         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1641         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1642         (*fxtractxf3_i387): Rename from *fxtractxf3.
1643         (fxtract_extend<mode>xf3_i387): New insn pattern.
1644         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1645         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1646         and gen_fyl2xxf3_i387().
1647
1648 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
1649
1650         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1651         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1652         get_loop_exit_edges.  Changed return type to bool.
1653         * cfgloop.h (fix_loop_placement): Declaration removed.
1654
1655 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1656
1657         * param.h (MIN_VECT_LOOP_BOUND): New.
1658         * params.def (MIN_VECT_LOOP_BOUND): New.
1659         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1660         argument - minimum threshold for number of iterations.
1661         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1662         argument to declaration.
1663         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1664         MIN_VECT_LOOP_BOUND.
1665         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1666         slpeel_tree_peel_loop_to_edge with additional argument.
1667         (vect_do_peeling_for_alignment): Likewise.
1668         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1669
1670 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1671
1672         PR target/30413
1673         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1674         operands of size 1.
1675
1676 2007-01-14  Jan Hubicka  <jh@suse.cz>
1677
1678         * tree-dfa.c (remove_referenced_var): New function.
1679         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1680         prune referenced vars list too.
1681         * tree-flow.h (remove_referenced_var): Declare.
1682
1683 2007-01-14  Jan Hubicka  <jh@suse.cz>
1684
1685         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1686         separately.
1687         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1688         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1689         * except.h (verify_eh_throw_table_statements): Kill prototype.
1690         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1691         into call exrepssions.
1692         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1693         eh trhow verifier.
1694         * tree-cfg.c: Include pointer-set.h.
1695         (verify_node_sharing): Work on pointer set.
1696         (verify_eh_throw_stmt_node): New.
1697         (verify_stmts): Use pointers sets, verify throw_stmt.
1698
1699 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1700
1701         * ipa-reference.c (analyze_function): Consider also addresses taken
1702         in phi nodes.
1703
1704 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1705
1706         * c-typeck.c (null_pointer_constant_p): Replace use of
1707         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1708         (build_c_cast): Likewise.
1709
1710 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1711
1712         * tree.h (force_fit_type_double): Remove unused final argument.
1713         * c-common.c (constant_expression_warning): Replace use of
1714         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1715         (convert_and_check): Likewise.
1716         (shorten_compare): Update call to force_fit_type_double.
1717         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1718         * convert.c (convert_to_pointer): Update call to
1719         force_fit_type_double.
1720         * fold-const.c (force_fit_type_double): Remove overflowed_const
1721         argument.
1722         (int_const_binop, fold_convert_const_int_from_int,
1723         fold_convert_const_int_from_real, fold_div_compare,
1724         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1725         fold_abs_const, fold_not_const): Remove the final argument from
1726         calls to force_fit_type_double.
1727
1728 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1729
1730         * configure.ac: Set insn to "nop" for spu-*-* also.
1731         * configure: Regenerate.
1732
1733 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
1734
1735         * builtins.def : Add BUILT_IN_FREE.
1736
1737 2007-01-12  Jan Hubicka  <jh@suse.cz>
1738
1739         PR tree-optimization/30443
1740         * tree-inline.c (tree_function_versioning): Do not optimize when
1741         cloning for inlining.
1742
1743 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
1744
1745         * doc/loop.texi: Document recording of loop exits.
1746         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1747         (update_single_exits_after_duplication,
1748         update_single_exit_for_duplicated_loop,
1749         update_single_exit_for_duplicated_loops): Removed.
1750         (duplicate_loop_to_header_edge): Do not call
1751         update_single_exits_after_duplication and
1752         update_single_exit_for_duplicated_loops.
1753         (loop_version): Do not update single_exit information.
1754         (fix_loop_structure): Use record_loop_exits instead of
1755         mark_single_exit_loops.
1756         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1757         the lists of loop exits.
1758         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1759         split_edge, merge_blocks): Update the lists of loop exits.
1760         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1761         loop_optimizer_init.
1762         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1763         of mark_single_exit_loops.
1764         (loop_optimizer_finalize): Call release_recorded_exits.
1765         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1766         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1767         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1768         update single exit information.
1769         * lambda-code.c (perfect_nestify): Ditto.
1770         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1771         (mark_single_exit_loops): Removed.
1772         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1773         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1774         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1775         functions.
1776         (get_loop_exit_edges, single_exit): Use recorded exit lists.
1777         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1778         (verify_loop_structure): Verify consistency of the exit lists.
1779         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
1780         (set_single_exit): Removed.
1781         * cfgloop.h (struct loop_exit): New function.
1782         (struct loop): single_exit_ field replaced by exits field.
1783         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1784         (struct loops): Added exits hash.
1785         (mark_single_exit_loops, set_single_exit): Declaration removed.
1786         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1787
1788 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1789
1790         * doc/invoke.texi: Avoid use of @headitem.
1791
1792 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1793
1794         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1795
1796 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1797
1798         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1799         list of targets.
1800
1801 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1802             Richard Sandiford  <richard@codesourcery.com>
1803
1804         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1805         * config/m68k/m68k.opt: Resort options.
1806         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1807
1808 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1809             Richard Sandiford  <richard@codesourcery.com>
1810
1811         * doc/invoke.texi: Document the macros that are defined by
1812         m68k's -mtune and -mhard-float options.
1813         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1814         (m68k_cpp_cpu_family): Likewise.
1815         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1816         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
1817         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
1818         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1819         (m68k_cpp_cpu_family): Likewise.
1820
1821 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1822
1823         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1824         macros besides mc68000 as tuning macros.  Use a switch statement
1825         to set them and mcpu32.
1826
1827 2007-01-12  Julian Brown  <julian@codesourcery.com>
1828
1829         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1830         (TARGET_68040_ONLY): Rename to...
1831         (TARGET_68040): ...this.
1832         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1833         * config/m68k/m68k.md: Likewise.
1834
1835 2007-01-12  Julian Brown  <julian@codesourcery.com>
1836             Nathan Sidwell  <nathan@codesourcery.com>
1837             Richard Sandiford  <richard@codesourcery.com>
1838
1839         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1840         the -mcpu= argument associated with the --with-cpu setting.
1841         Define M68K_DEFAULT_TUNE to the default -mtune= option,
1842         if different from the one implied by the -mcpu setting.
1843         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1844         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
1845         * doc/invoke.texi: Mention ColdFire in the introduction to the
1846         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
1847         -mno-div and -mhard-float options.  Make -m68881 a synonym for
1848         -mhard-float.  Document the previously-undocumented -m5206e,
1849         -m528x, -m5307 and -m5407 options.  Tweak the existing option
1850         documentation for consistency.
1851         * doc/install.texi: Mention new --with-cpu arguments.
1852         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1853         default CPU if neither -mcpu nor -march are specified.
1854         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1855         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1856         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
1857         using m68k_tune to decide between families that implement the
1858         same ISA.  Use m68k_tune to set __mcfv4e__.
1859         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1860         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1861         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1862         (FL_ISA_C, FL_ISA_MMU): New macros.
1863         (MASK_COLDFIRE): Delete.
1864         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1865         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1866         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1867         (TARGET_HARD_FLOAT): Do not define here.
1868         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1869         (TUNE_68000): New macro.
1870         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1871         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1872         (TUNE_CFV2): Redefine in terms of m68k_tune.
1873         (uarch_type, target_device, fpu_type): New enums.
1874         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1875         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1876         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1877         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1878         (FL_FOR_isa_c): New macros.
1879         (m68k_isa): New enum.
1880         (m68k_target_selection): New structure.
1881         (all_devices, all_isas, all_microarchs): New tables.
1882         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1883         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1884         (MASK_ALL_CPU_BITS): Delete.
1885         (m68k_find_selection): New function.
1886         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1887         Map the legacy target options to a combination of the new ones.
1888         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1889         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
1890         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1891         settings.
1892         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1893         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1894         properties.
1895         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1896         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1897         * config/m68k/m68k-devices.def: New file.
1898
1899 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1900             Nathan Sidwell  <nathan@codesourcery.com>
1901
1902         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1903         (SUBTARGET_EXTRA_SPECS): New macros.
1904         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1905         use %(asm_cpu_spec) instead.
1906         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1907         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1908         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1909         (EXTRA_SPECS): Rename to...
1910         (SUBTARGET_EXTRA_SPECS): ...this.
1911
1912 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1913             Richard Sandiford  <richard@codesourcery.com>
1914             Julian Brown  <julian@codesourcery.com>
1915
1916         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1917         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1918         to the configuration's default CPU.
1919         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1920         Remove default masks.
1921         (m680[012]0-*-*): Set the default with_cpu to the first part of
1922         the target name.
1923         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1924         (m68k*-*-linux): Extend the --with-cpu handling to...
1925         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
1926         and m68010.  Don't set target_cpu_default2.
1927         * doc/install.texi: Document --with-cpu for m68k.
1928         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1929         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1930         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1931         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1932         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1933         (ASM_SPEC): Remove use of %(asm_cpu_default).
1934         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1935         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1936         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
1937         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1938         __HAVE_FPU__ if TARGET_HARD_FLOAT.
1939         (TARGET_DEFAULT): Delete.
1940         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1941         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1942         (CPP_CPU_SPEC): Delete.
1943         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1944         the appropriate string.
1945         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1946         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1947         (ASM_SPEC): Don't use %(asm_default_spec).
1948         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1949         TARGET_DEFAULT and add MASK_68881.
1950         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1951
1952 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1953
1954         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1955         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1956         MASK_68010 alongside MASK_68020.
1957         * doc/invoke.texi: Document -m68010.
1958         * config/m68k/m68k.opt (m68010): New.
1959         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1960         if TUNE_68010.
1961         (TUNE_68010): New macro.
1962         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1963         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1964         (M68K_CPU_m68332): Add MASK_68010.
1965         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1966         fallback definition.
1967         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1968         defines.
1969         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1970         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
1971         to all entries that use MASK_68020.
1972         (output_move_simode_const, output_move_himode, output_move_qimode)
1973         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1974         instead of TARGET_68020 to select clr behavior.  Remove comment
1975         about there being no TARGET_68010.
1976         * config/m68k/m68k.md: Likewise throughout.
1977
1978 2007-01-12  Julian Brown  <julian@codesourcery.com>
1979
1980         * config/m68k/m68k.h (TARGET_ISAB): New macro.
1981         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1982         * config/m68k/m68k.md: Likewise.
1983
1984 2007-01-12  Julian Brown  <julian@codesourcery.com>
1985
1986         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1987         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1988
1989 2007-01-12  Julian Brown  <julian@codesourcery.com>
1990
1991         * config/m68k/m68k.h (TUNE_68040_60): New macro.
1992         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1993         * config/m68k/m68k.md: Likewise.
1994
1995 2007-01-12  Julian Brown  <julian@codesourcery.com>
1996             Richard Sandiford  <richard@codesourcery.com>
1997
1998         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1999         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2000         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2001         TARGET_CPU32.
2002         (TARGET_CPU32): Rename to...
2003         (TUNE_CPU32): ...this.
2004         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2005         (TUNE_CFV2): New macros.
2006         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2007         remove conditions that are implied by TARGET_68020.
2008         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2009         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2010         (m68k_output_function_epilogue): Likewise.
2011         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2012         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2013         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2014         non-68000 timings.  Refactor multiplication and division costs.
2015         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2016         TUNE_CPU32 instead of TARGET_CPU32.
2017         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2018         and TUNE_68060 instead of TARGET_68060.
2019         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2020         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2021         TARGET_CPU32.
2022         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2023         && !TARGET_COLDFIRE" to choose between moveq and clr.
2024         Likewise in the unnamed movsf pattern.
2025         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2026         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2027         ashiftrt pattern.
2028
2029 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2030
2031         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2032         of tabbing before backslashes.
2033
2034 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2035
2036         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2037         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2038         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2039
2040 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2041
2042         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2043         find_interesting_uses_cond.
2044         (find_interesting_uses_cond): Use extract_cond_operands.
2045         (rewrite_use_compare): Use extract_cond_operands and
2046         force_gimple_operand_bsi.  Do not call update_stmt.
2047         (determine_use_iv_cost_condition): Use extract_cond_operands.
2048         Return cheaper of using original bound and changing the exit bound.
2049
2050 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2051
2052         PR tree-optimization/29516
2053         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2054         most_expensive_mult_to_index, addr_to_parts,
2055         create_mem_ref, maybe_fold_tmr): Make the type of
2056         fields of TARGET_MEM_REF sizetype.
2057         (move_fixed_address_to_symbol, move_pointer_to_base):
2058         New functions.
2059         * tree.def (TARGET_MEM_REF): Add comment on types of
2060         the operands.
2061
2062 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
2063
2064         * c-common.c (vector_types_convertible_p): Treat opaque types as
2065         always convertible if they have the same size, but not otherwise.
2066
2067 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
2068
2069         * ifcvt.c (struct noce_if_info): Add comments to the fields.
2070         Remove the b_unconditional field.
2071         (noce_try_sign_mask): Do not look at b_unconditional.
2072         (noce_process_if_block): Do not use merge_if_blocks.  Update
2073         the CFG here.  Do not set b_unconditional.
2074         (cond_move_process_if_block): Likewise.
2075         (find_cond_trap): Likewise.
2076         (check_cond_move_block): Require simple jump insns at the end
2077         of the basic block.
2078
2079 2007-01-11  Jan Hubicka  <jh@suse.cz>
2080
2081         PR tree-optimization/1046
2082         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2083         when alias info is not ready.
2084         (pass_tail_recursion): Do not require aliasing.
2085         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2086         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2087         * tree-ssa-copy.c (pass_copy_prop): Likewise.
2088         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2089         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2090         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2091         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2092         before inlining.
2093         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2094         When aliasing is not build, mark statement as volatile.
2095
2096 2007-01-11  Tom Tromey  <tromey@redhat.com>
2097
2098         PR preprocessor/15185, PR preprocessor/20989:
2099         * doc/cppopts.texi <-MT>: Update description of algorithm for
2100         computing default target.
2101         <-M, -MD>: Reword "basename" text.
2102
2103 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2104
2105         * builtins.c (expand_builtin_pow, expand_builtin_powi,
2106         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2107         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2108         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2109         fold_builtin_bswap, real_constp, fold_builtin_pow,
2110         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2111         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2112         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2113         * convert.c (convert_to_pointer): Likewise.
2114         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2115         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2116         const_binop, fold_convert_const_int_from_int,
2117         fold_convert_const_int_from_real,
2118         fold_convert_const_real_from_real, sign_bit_p,
2119         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2120         fold_sign_changed_comparison, fold_unary, fold_comparison,
2121         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2122         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2123         * print-tree.c (print_node_brief, print_node): Likewise.
2124         * stor-layout.c (place_field, layout_type): Likewise.
2125         * tree-chrec.c (keep_cast): Likewise.
2126         * tree.c (build_vector, build_real, build_real_from_int_cst,
2127         build_complex): Likewise.
2128
2129 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2130
2131         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
2132         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2133
2134 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
2135
2136         * configure.ac (strict1_warn): Rename to strict_warn.
2137         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2138         cc_set_by_configure, quoted_cc_set_by_configure,
2139         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2140         all_boot_languages, all_stagestuff): Remove.
2141         (target_list): Remove bootstrap targets.
2142         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2143         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2144         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2145         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2146         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2147         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2148         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2149         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2150         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2151         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2152         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2153         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2154         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2155         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2156         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2157         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2158         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2159         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2160         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2161         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2162         (mostlyclean): Adjust.
2163         (clean, distclean): Don't mention bootstrap stuff.
2164         * configure: Regenerate.
2165         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2166         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2167         treelang/config-lang.in (stagestuff): Remove.
2168         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2169         stageprofile, stagefeedback, stagestuff): Remove mention.
2170
2171 2007-01-11  Nick Clifton  <nickc@redhat.com>
2172
2173         * config/mcore/predicates.md (mcore_general_movesrc_operand):
2174         Accept CONSTs.
2175         (mcore_general_movdst_operand): Do not accept CONST_INTs.
2176         (mcore_arith_K_S_operand): Run the test for the S constraint not
2177         the test for the M constraint.
2178         (mcore_addsub_operand): Do not accept integer values that are
2179         larger than 32 bits.
2180         * config/mcore/mcore.md: Remove unused constraints from split.
2181         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2182         (addsi3): Likewise.
2183         (allocate_stack): Likewise.
2184         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2185         operands to 32 bits.
2186         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2187         INTVAL.
2188         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2189         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2190         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2191         mcore_output_bclri, mcore_output_andn, output_inline_const,
2192         mcore_output_move, mcore_output_movedouble): Likewise.
2193         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2194         (output_inline_const): Likewise.
2195         (output_inline_const): Fix format strings used in sprintf
2196         statements.
2197         * config/mcore/mcore-protos.h: Update prototypes for changed
2198         functions in mcore.c.
2199         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2200         HOST_WIDE_INT and not int.
2201         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2202         CONST_OK_FOR_N): Likewise.
2203         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2204         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2205         an INTVAL.
2206
2207 2007-01-10  Jan Hubicka  <jh@suse.cz>
2208
2209         * tree-vrp.c (remove_range_assertions): Release defs.
2210         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2211         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2212
2213 2007-01-10  Paul Brook  <paul@codesourcery.com>
2214
2215         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2216         * config/arm/arm.md (mulsi3subsi): New insn.
2217
2218 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
2219
2220         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2221         (tree_transform_and_unroll_loop): New.
2222         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2223         Declare.
2224
2225 2007-01-10  Robert Kennedy <jimbob@google.com>
2226
2227         * fold-const.c (fold_comparison): Fold comparisons like (x *
2228         1000 < 0) to (x < 0).
2229
2230 2007-01-10  Ian Lance Taylor  <iant@google.com>
2231
2232         * tree-pretty-print.c (dump_generic_node): Print parentheses when
2233         operands have the same priority.
2234
2235 2007-01-10  Tom Tromey  <tromey@redhat.com>
2236
2237         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2238         (fold_binary): Likewise.
2239         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2240         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2241         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2242         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2243         * langhooks.h (struct lang_hooks): Removed field
2244         'can_use_bit_fields_p'.
2245
2246 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
2247
2248         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2249
2250 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
2251
2252         * function.c (get_last_funcdef_no): New function.
2253         * function.h (get_last_funcdef_no): Declare.
2254         * tree-inline.c (initialize_cfun): Add initialization.
2255         (tree_function_versioning): Cleanup.
2256
2257 2007-01-10  Jan Hubicka  <jh@suse.cz>
2258
2259         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2260         PHIs.
2261
2262 2007-01-10  Sa Liu  <saliu@de.ibm.com>
2263             Ben Elliston  <bje@au.ibm.com>
2264
2265         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2266         for SAVE_NONLOCAL and Pmode for any other save level.
2267         * spu-protos.h (spu_restore_stack_block): Declare.
2268         * spu.md (save_stack_block): Remove.
2269         (restore_stack_block): Call spu_restore_stack_block.
2270         * spu.c (spu_restore_stack_block): New function.
2271         (spu_expand_epilogue): Remove old comment.
2272
2273 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
2274
2275         PR tree-optimization/30322
2276         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2277         (cand_value_at): Return the value as aff_tree.
2278         (may_eliminate_iv): Convert the bound from aff_tree to tree.
2279         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2280         aff_combination_mult): New functions.
2281         (aff_combination_add): Use aff_combination_add_cst.
2282         (aff_combination_convert): Allow conversions to a wider type.
2283         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2284         * tree-affine.h (aff_combination_mult): Declare.
2285
2286 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
2287
2288         * doc/tm.texi: Update documentation to reflect reality of exec
2289         and start file search behaviours. Update copyright year.
2290         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2291         header file directories.
2292
2293 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
2294
2295         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2296         (*cosxf2): Rename to cosxf2_i387.
2297         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
2298         using X87MODEF12 mode macro. Rename patterns to
2299         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
2300         for SSE math.
2301         (*cosdf2, *cossf2): Ditto.
2302         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
2303         to match extended input operands.
2304         (sincos<mode>3): New expander.
2305         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2306         insn patterns and corresponding splitters.
2307
2308 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
2309
2310         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2311         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2312
2313 2007-01-09  Nicolas Pitre  <nico@cam.org>
2314
2315         PR target/30173
2316         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2317
2318 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2319
2320         * target.h (struct gcc_target): New field library_rtti_comdat.
2321         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2322         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2323         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2324         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2325
2326 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2327
2328         * doc/invoke.texi (Optimize Options): Correct description of -O0.
2329
2330 2007-01-08  Richard Guenther  <rguenther@suse.de>
2331
2332         * tree.h (force_fit_type_double): Export.
2333         (force_fit_type): Remove.
2334         * fold-const.c (force_fit_type_double): New function.
2335         (force_fit_type): Remove.
2336         (int_const_binop): Use it.
2337         (fold_convert_const_int_from_int): Likewise.
2338         (fold_convert_const_int_from_real): Likewise.
2339         (fold_div_compare): Likewise.
2340         (fold_sign_changed_comparison): Likewise.
2341         (fold_unary): Likewise.
2342         (fold_negate_const): Likewise.
2343         (fold_abs_const): Likewise.
2344         (fold_not_const): Likewise.
2345         * c-common.c (shorten_compare): Use force_fit_type_double.
2346         * convert.c (convert_to_pointer): Likewise.
2347
2348 2007-01-08  Richard Guenther  <rguenther@suse.de>
2349
2350         * tree.h (build_int_cst_wide_type): Export.
2351         * tree.c (build_int_cst_wide_type): New function.
2352         (build_int_cst_wide): Fix comment.
2353         * builtins.c (fold_builtin_object_size): Use build_int_cst
2354         to build -1 or 0 of the correct type.  Use fit_double_type
2355         to check for overflow.
2356         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2357         to build the mask.
2358         (decode_field_reference): Likewise.
2359         (all_ones_mask_p): Likewise.
2360         (native_interpret_int): Use build_int_cst_wide_type.
2361         (fold_binary): Use build_int_cst_type to build an all-ones
2362         value.
2363         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2364
2365 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
2366
2367         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2368
2369 2007-01-09  Ben Elliston  <bje@au.ibm.com>
2370
2371         * genautomata.c (STATS_OPTION): New option.
2372         (stats_flag): New flag.
2373         (gen_automata_option): Handle it.
2374         (initiate_automaton_gen): Ditto.
2375         (write_automata): Output statistics only if stats_flag is
2376         set. Likewise, output time statistics only if time_flag is set.
2377         * doc/md.texi (Processor pipeline description): Document new flag.
2378
2379 2007-01-08  Richard Guenther  <rguenther@suse.de>
2380
2381         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2382         * tree.c (build_int_cst_type): Likewise.
2383         (size_in_bytes): Don't call force_fit_type on the result.
2384         (int_fits_type_p): Use fit_double_type.
2385         * fold-const.c (fit_double_type): New function.
2386         (force_fit_type): Use it.
2387         * tree.h (fit_double_type): Export.
2388
2389 2007-01-08  Jan Hubicka  <jh@suse.cz>
2390
2391         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2392         * ipa.c (function_and_variable_visibility): Fix return type.
2393
2394 2007-01-08  Richard Guenther  <rguenther@suse.de>
2395
2396         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2397         of offset to build the index.
2398         * tree-pretty-print.c (dump_generic_node): Don't build negated
2399         const just for printing.
2400         * c-pretty-print.c (pp_c_integer_constant): Likewise.
2401         * builtins.c (fold_builtin_int_roundingfn): Check if result
2402         fits the type by using force_fit_type and comparing the result.
2403         * predict.c (predict_loops): Use compare_tree_int for comparison.
2404         * tree.c (build_int_cst): Fall back to integer_type_node for
2405         NULL_TREE type.
2406         (build_int_cst_wide): Assert type is non-null.
2407
2408 2007-01-08  Roberto Costa  <roberto.costa@st.com>
2409
2410         * tree-vrp.c (extract_range_from_cond_expr): New.
2411         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2412         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2413         as expressions.
2414         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2415         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2416         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2417         used as expressions.
2418         * tree-object-size.c (cond_expr_object_size): New.
2419         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2420
2421 2007-01-08  Jan Hubicka  <jh@suse.cz>
2422
2423         * tree-ssa-forwprop.c (forward_propagate_into_cond,
2424         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2425         statement.
2426
2427 2007-01-08  Richard Guenther  <rguenther@suse.de>
2428
2429         PR tree-optimization/23603
2430         * tree-vrp.c (set_value_range_to_truthvalue): New function.
2431         (extract_range_from_binary): Fall back to truthvalue instead of
2432         varying for TRUTH_*_EXPR.
2433         (extract_range_from_comparison): Fall back to truthvalue instead of
2434         varying.
2435         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2436         if all visited PHI values were constant.
2437
2438 2007-01-08  Jan Hubicka  <jh@suse.cz>
2439
2440         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2441         (cgraph_analyze_function): break out from ...
2442         (cgraph_finalize_compilation_unit): ... here.
2443         (cgraph_expand_function): Remove forgoten commented out line.
2444         (cgraph_optimize): Analyze functions.
2445
2446 2007-01-08  Jan Hubicka  <jh@suse.cz>
2447
2448         * tree-pas.h (TODO_remove_function): New flag.
2449         (TODO_update*): Renumber.
2450         (pass_ipa_increase_alignment,
2451         pass_ipa_function_and_variable_visibility): New passes.
2452         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2453         (cgraph_function_and_variable_visibility): Move to ipa.c
2454         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2455         cgraph_increase_alignment.
2456         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2457         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2458         calling tree-inline.
2459         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2460         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2461         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2462         (gate_increase_alignment): New function.
2463         (pass_ipa_increase_alignment): New pass.
2464         * ipa.c: Inline tree-pass.h and timevar.h
2465         (function_and_variable_visibility): Move here from cgraphunit.c
2466         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2467         * passes.c (init_optimization_passes): Add the two new passes.
2468         (execute_todo): Handle cgraph_remove_functions.
2469
2470 2007-01-08  Nick Clifton  <nickc@redhat.com>
2471
2472         * config/frv/predicates.md (reg_or_0_operand): Accept
2473         CONST_DOUBLEs.
2474
2475 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
2476
2477         * config/bfin/rtems.h, config/bfin/t-rtems: New.
2478         * config.gcc: Add bfin*-rtems*.
2479
2480 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2481
2482         * c.opt: Add -flax-vector-conversions.
2483         * c-typeck.c (convert_for_assignment): Pass flag to
2484         vector_types_convertible_p to allow emission of note.
2485         (digest_init): Likewise.
2486         * c-opts.c: Handle -flax-vector-conversions.
2487         * c-common.c (flag_lax_vector_conversions): New.
2488         (vector_types_convertible_p): Unless -flax-vector conversions
2489         has been passed, disallow conversions between vectors with
2490         differing numbers of subparts and/or element types.  If such
2491         a conversion is disallowed, possibly emit a note on the first
2492         occasion only to inform the user of -flax-vector-conversions.
2493         The new last argument specifies this.
2494         * c-common.h (flag_lax_vector_conversions): New.
2495         (vector_types_convertible_p): Add extra argument.
2496         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2497         char_type_node for V*QI type vectors.
2498         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2499         Update to satisfy new typechecking rules.
2500         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2501         C and C++ variants.
2502         * doc/invoke.texi (C Dialect Options): Document
2503         -flax-vector-conversions.
2504
2505 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2506
2507         PR tree-optimization/29877
2508         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2509         a register variable on the RHS to not be replaceable.
2510
2511 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
2512         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2513         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2514         const_simm15, const_pow2, const_npow2): Added.
2515         * config/score/misc.md (insv, extv, extzv, movmemsi,
2516         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2517         * config/score/score.c (score_address_cost, score_select_cc_mode):
2518         Added.
2519         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2520         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2521         Update score_print_operand.
2522         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2523         Adjust register allocate order and update some macro define.
2524         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2525         mdx_block_move_straight, mdx_block_move_loop_head,
2526         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2527         mdx_block_move): Added.
2528         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2529         mdp_limm): Updated and fix some bug and typo.
2530         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2531         (movsf, movdf, doloop_end): Added.
2532
2533 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
2534
2535         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2536         config/arm/thumb2.md: Fix comment typos.
2537         * doc/extend.texi: Fix a typo.
2538
2539 2007-01-07  Eric Christopher  <echristo@apple.com>
2540
2541         * configure.ac: Check for __stack_chk_fail for darwin.
2542         * configure: Regenerate.
2543
2544 2007-01-07  Richard Guenther  <rguenther@suse.de>
2545
2546         * tree-vrp.c (extract_range_from_assert): CSE calls to
2547         compare_values where possible.
2548         (extract_range_from_unary_expr): Likewise.
2549
2550 2007-01-07  Anatoly Sokolov <aesok@post.ru>
2551
2552         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2553         * config/avr/avr.c (call_insn_operand): Delete function.
2554         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2555         instead of register number. Use predicates.
2556         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2557         call_insn_operand): Add.
2558
2559 2007-01-06  Jan Hubicka  <jh@suse.cz>
2560
2561         * tree-pass.h (pass_build_cgraph_edges): Declare.
2562         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2563         (visited_nodes): Remove.
2564         (cgraph_create_edges): Move to cgraphbuild.c; rename to
2565         build_cgrpah_edges; make visited_nodes local.
2566         (cgraph_process_new_functions): DO not call initialize_inline_failed.
2567         (record_references_in_initializer): Move to cgraphbuild.c
2568         (initialize_inline_failed, rebuild_cgraph_edges,
2569         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2570         (verify_cgraph_node): Make visited_nodes local.
2571         (cgraph_analyze_function): Do not call cgraph_create_edges and
2572         initialize_inline_failed.
2573         (cgraph_expand_function): Do not call cgraph_lower_function;
2574         assert that function is already lowered.
2575         * Makefile.in (cgraphbuild.o): New.
2576         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2577         at the end of lowering passes.
2578
2579 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
2580
2581         * ifcvt.c (cond_move_convert_if_block): New function, code
2582         factored out from...
2583         (cond_move_process_if_block): ...here.  Call the new function
2584         on the THEN and ELSE blocks.
2585         (merge_if_block): Do not copy global_live_at_end, merge_blocks
2586         already takes care of this.
2587
2588 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2589
2590         PR c/19978
2591         * tree.h (TREE_OVERFLOW_P): New.
2592         * c-typeck.c (parser_build_unary_op): Warn only if result
2593         overflowed and operands did not.
2594         (parser_build_binary_op): Likewise.
2595         (convert_for_assignment): Remove redundant overflow_warning.
2596         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2597
2598 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2599
2600         * c-typeck.c (store_init_value): Split over two lines to follow
2601         the GNU coding style.
2602
2603 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
2604
2605         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2606         __GXX_EXPERIMENTAL_CXX0X__.
2607         * doc/cpp.texi: Same.
2608
2609 2007-01-05  Richard Guenther  <rguenther@suse.de>
2610
2611         PR middle-end/27826
2612         * tree.c (get_narrower): Do not construct COMPONENT_REFs
2613         with mismatched types.  Instead explicitly build a
2614         conversion NOP_EXPR.
2615
2616 2007-01-05  Ian Lance Taylor  <iant@google.com>
2617
2618         * c-common.c (decl_with_nonnull_addr_p): New function.
2619         (c_common_truthvalue_conversion): Call it.
2620         * c-typeck.c (build_binary_op): Likewise.
2621         * c-common.h (decl_with_nonnull_addr_p): Declare.
2622
2623 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
2624
2625         PR c/30360
2626         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2627         denom against 0.0.
2628
2629 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
2630
2631         * doc/install.texi (Final install): Document the fact that
2632         the GNAT runtime should not be stripped.
2633
2634 2007-01-04  Jan Hubicka  <jh@suse.cz>
2635
2636         * tree-inline.c (fold_marked_statements): Update operand caches
2637         and EH after folding
2638
2639 2007-01-04  Ian Lance Taylor  <iant@google.com>
2640
2641         * c-common.c (check_function_nonnull): Whitespace fix.
2642
2643 2007-01-04  Jan Hubicka  <jh@suse.cz>
2644
2645         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2646         comitted older version of patch.
2647         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2648         function, celanup cfg and collect garbage.
2649
2650 2007-01-04  Mike Stump  <mrs@apple.com>
2651
2652         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2653         (clean): Likewise.
2654
2655 2007-01-04  Eric Christopher  <echristo@apple.com>
2656
2657         * libgcc2.c (__bswapsi2): Use SItype.
2658         (__bswapdi2): Use DItype.
2659         * libgcc2.h: Update for above.
2660
2661 2007-01-04  Paul Brook  <paul@codesourcery.com>
2662
2663         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2664         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2665         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2666         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2667         umulsi3_highpart): Make conditional on !arm_arch6.
2668         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2669         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2670         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2671         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2672         umulsi3_highpart_v6): New insns.
2673
2674 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2675
2676         * fold-const.c (fold_convert): When casting an expression to void,
2677         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2678         have a type.  Instead of attempting to build a NOP_EXPR, return
2679         these "special" trees directly.
2680
2681 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2682
2683         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2684         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2685         rs6000_cost->fp not 0 as adjustment for outer NEG.
2686
2687 2007-01-04  Jan Hubicka  <jh@suse.cz>
2688
2689         * cgraph.c (cgraph_release_function_body): New function.
2690         (cgraph_remove_node): Use it.
2691         * cgraph.h (cgraph_release_function_body): Declare.
2692         * cgraphunit.c (cgraph_expand_function): Use it.
2693         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2694         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2695         is initialized and while compilation of other function is running.
2696         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2697         statement CFG annotations from here to ...
2698         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2699
2700 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2701
2702         * cfgloop.h (enum li_flags): Make the constants powers of two.
2703
2704 2007-01-04  Jan Hubicka  <jh@suse.cz>
2705
2706         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2707         set.
2708         (fold_marked_statements): New function.
2709         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2710         * tree-inline.h (copy_body_data): Add statements_to_fold.
2711
2712 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2713
2714         * config.gcc: Mention libgcc/config.host.
2715         * Makefile.in: Update comments mentioning libgcc.
2716         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2717         (all.cross, start.encap, rest.encap, rest.cross): Update
2718         dependencies for libgcc move.
2719         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2720         (clean-target, clean-target-libgcc): Delete.
2721         (srcdirify, GCC_EXTRA_PARTS): New macros.
2722         (libgcc-support, libgcc.mvars): New rules.
2723         (distclean): Remove mention of mklibgcc.
2724         (install): Don't reference INSTALL_LIBGCC.
2725         (install-common): Don't reference EXTRA_PARTS.
2726         (install-libgcc, install-multilib): Delete rules.
2727         * mklibgcc.in: Delete file.
2728         * doc/configfiles.texi: Don't mention mklibgcc.
2729
2730         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2731         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2732         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2733         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2734         (mklibgcc, ldblspecs): Likewise.
2735
2736         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2737         (posixpre.def): Use $(T).
2738         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2739         (SHLIB_SRC, SHLIB_INSTALL): Delete.
2740         (SHLIB_LINK): Make dummy.
2741         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2742
2743         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2744
2745         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2746         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2747         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2748         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2749
2750         * configure.ac (all_outputs): Remove mklibgcc.
2751         * configure: Regenerated.
2752
2753 2007-01-03  Josh Conner  <jconner@apple.com>
2754
2755         PR middle-end/29683
2756         * calls.c (compute_argument_addresses): Set stack and stack_slot
2757         for partial args, too.
2758         (store_one_arg): Use locate.size.constant for the size when
2759         generating a save_area.
2760
2761 2007-01-03  Robert Kennedy <jimbob@google.com>
2762
2763         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2764         whose definitions are deleted due to basic block merging.
2765
2766 2007-01-03  Paul Brook  <paul@codesourcery.com>
2767
2768         PR target/16634
2769         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2770         functions.
2771         (use_return_insn): Return 0 for Thumb interrupt functions.
2772         (print_multi_reg): Add rfe argument for IRQ returns.
2773         (arm_output_epilogue): Pop interrupt return address directly into PC.
2774         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2775
2776 2007-01-03  Paul Brook  <paul@codesourcery.com>
2777
2778         Merge from sourcerygxx-4_1.
2779         * config/arm/thumb2.md: New file.
2780         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2781         Thumb-2.
2782         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2783         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2784         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2785         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2786         tables.
2787         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2788         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2789         comptibility.
2790         * config/arm/ieee754-sf.S: Ditto.
2791         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2792         (thumb1_base_register_rtx_p): ... to this.
2793         (thumb_index_register_rtx_p): Rename...
2794         (thumb1_index_register_rtx_p): ... to this.
2795         (thumb_output_function_prologue): Rename...
2796         (thumb1_output_function_prologue): ... to this.
2797         (thumb_legitimate_address_p): Rename...
2798         (thumb1_legitimate_address_p): ... to this.
2799         (thumb_rtx_costs): Rename...
2800         (thumb1_rtx_costs): ... to this.
2801         (thumb_compute_save_reg_mask): Rename...
2802         (thumb1_compute_save_reg_mask): ... to this.
2803         (thumb_final_prescan_insn): Rename...
2804         (thumb1_final_prescan_insn): ... to this.
2805         (thumb_expand_epilogue): Rename...
2806         (thumb1_expand_epilogue): ... to this.
2807         (arm_unwind_emit_stm): Rename...
2808         (arm_unwind_emit_sequence): ... to this.
2809         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2810         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2811         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2812         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2813         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2814         thumb2_output_casesi): New functions.
2815         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2816         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2817         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2818         THUMB2_WORK_REGS): Define.
2819         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2820         arm_condexec_mask, arm_condexec_masklen)): New variables.
2821         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2822         (arm_override_options): Check new CPU capabilities.
2823         Set new architecture flag variables.
2824         (arm_isr_value): Handle v7m interrupt functions.
2825         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
2826         Thumb-2.
2827         (const_ok_for_arm): Handle Thumb-2 constants.
2828         (arm_gen_constant): Ditto.  Use movw when available.
2829         (arm_function_ok_for_sibcall): Return false for v7m interrupt
2830         functions.
2831         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2832         (thumb_find_work_register, arm_load_pic_register,
2833         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2834         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2835         print_multi_reg, output_mov_long_double_fpa_from_arm,
2836         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2837         output_mov_double_fpa_from_arm, output_move_double,
2838         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2839         output_return_instruction, arm_output_function_prologue,
2840         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2841         arm_output_mi_thunk, thumb_set_return_address): Ditto.
2842         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
2843         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2844         (arithmetic_instr, shift_op): Use arm_shift_nmem.
2845         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
2846         '!' and 'L'.
2847         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2848         (thumb_expand_prologue): Use thumb_set_frame_pointer.
2849         (arm_file_start): Output directive for unified syntax.
2850         (arm_unwind_emit_set): Handle stack alignment instruction.
2851         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2852         Add v6t2, v7, v7a, v7r and v7m.
2853         (RETLDM): Add Thumb-2 code.
2854         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2855         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2856         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2857         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2858         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2859         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2860         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2861         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2862         ADJUST_INSN_LENGTH): Define.
2863         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2864         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2865         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2866         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2867         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2868         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2869         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2870         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2871         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2872         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2873         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2874         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2875         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2876         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2877         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2878         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2879         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2880         cortex-m3.
2881         * config/arm/arm-tune.md: Regenerate.
2882         * config/arm/arm-protos.h: Update prototypes.
2883         * config/arm/vfp.md: Enable patterns for Thumb-2.
2884         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
2885         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2886         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2887         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2888         * config/arm/libunwind.S: Add Thumb-2 code.
2889         * config/arm/constraints.md: Update include Thumb-2.
2890         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2891         * config/arm/ieee754-df.S: Ditto.
2892         * config/arm/bpabi.S: Ditto.
2893         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2894         * config/arm/predicates.md (low_register_operand,
2895         low_reg_or_int_operand, thumb_16bit_operator): New.
2896         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2897         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2898         * config/arm/t-arm-elf: Add armv7 multilib.
2899         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2900         Include thumb2.md.
2901         (UNSPEC_STACK_ALIGN, ce_count): New.
2902         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2903         arm_zero_extendsidi2, arm_zero_extendqidi2): New
2904         insns/expanders.
2905         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2906         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2907         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2908         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2909         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2910         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2911         insns.
2912         * doc/extend.texi: Document ARMv7-M interrupt functions.
2913         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2914
2915 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
2916
2917         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2918         (struct _Unwind_Context): Rename args_size to flags, remove
2919         signal_frame field, add a new args_size field and version field.
2920         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2921         _Unwind_IsExtendedContext): New inline functions.
2922         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2923         Assume by_value array is only present if _Unwind_IsExtendedContext.
2924         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2925         _Unwind_IsSignalFrame.
2926         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2927         (uw_update_context_1): Use _Unwind_SetSignalFrame.
2928         (uw_init_context_1): Initialize context->flags to
2929         EXTENDED_CONTEXT_BIT.
2930         * config/rs6000/linux-unwind.h (frob_update_context): Use
2931         _Unwind_SetSignalFrame.
2932
2933 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
2934
2935         PR middle-end/30353
2936         * gimplify.c (gimplify_modify_expr_complex_part): Move below
2937         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
2938         the value.
2939
2940 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
2941
2942         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2943         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2944         RET, AT_SP): Remove.
2945         * config/i386/i386.md (*sse_prologue_save_insn): Use return
2946         instead of RET.
2947
2948         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2949         * cfgbuild.c (find_basic_blocks): Likewise.
2950         * cfgrtl.c (rtl_create_basic_block): Likewise.
2951         * function.c (temp_slots_at_level): Likewise.
2952         * reg-stack.c (stack_regs_mentioned): Likewise.
2953         * regclass.c (allocate_reg_info): Likewise.
2954         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2955         set_bb_for_stmt, move_block_to_fn): Likewise.
2956         * tree-complex.c (tree_lower_complex): Likewise.
2957         * vec.h (VEC_safe_grow_cleared): New.
2958
2959         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2960
2961 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
2962
2963         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2964         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2965         to loop_version.
2966         * cfgloopmanip.c (scale_loop_frequencies): Export.
2967         (loopify): Scale the frequencies by prescribed coefficients.
2968         (set_zero_probability): New function.
2969         (duplicate_loop_to_header_edge): Improve updating of frequencies.
2970         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2971         and frequencies according to arguments.
2972         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2973         correctly.
2974         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2975         * modulo-sched.c (sms_schedule): Set probabilities for entering
2976         versioned loop correctly.
2977         * tree-vect-transform.c (vect_transform_loop): Ditto.
2978         * cfgloop.h (loopify, loop_version): Declaration changed.
2979         (scale_loop_frequencies): Declared.
2980
2981 2007-01-02  Jan Hubicka  <jh@suse.cz>
2982
2983         * cgraph.c: Include tree-flow.h
2984         (cgraph_add_new-function): Handle IPA_SSA mode; execute
2985         early_local_passes.
2986         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2987         * tree-pass.h (pass_all_early_optimizations): Declare.
2988         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2989         early_local_passes.
2990         (cgraph_analyze_function): Do early_local_passes.
2991         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2992         Do not add referenced vars.
2993         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2994         SSA form.
2995         (gate_all_early_local_passes): New gate.
2996         (pass_early_local_passes): Use new gate.
2997         (execute_early_local_optimizations): New functions.
2998         (gate_all_early_optimizations): New gate.
2999         (pass_all_early_optimizations): New pass.
3000         (execute_free_datastructures): Free SSA only when initialized.
3001         (gate_init_datastructures): Init only when optimizing.
3002         (tree_lowering_passes): Do early local passes when called late.
3003         * tree-profile.c (do_tree_profiling): Don't profile functions added
3004         late.
3005         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3006         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3007         not active.
3008         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3009         during early_local_passes.
3010         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3011
3012
3013 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3014
3015         * Makefile.in: Update copyright year.
3016
3017 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3018
3019         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3020         in $(lang_checks) and check-consistency targets.
3021
3022 2007-01-02  Jan Hubicka  <jh@suse.cz>
3023
3024         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3025         Do not add referenced vars.
3026         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3027         are not active.
3028         * passes.c (init_optimization_passes): Put mudflap_2 after
3029         free_datastructures.
3030
3031 2007-01-02  Jan Hubicka  <jh@suse.cz>
3032
3033         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3034         Set NOTHROW flag on call statements proved to be nothrow.
3035         Update statement of local calls so new pure/const functions are
3036         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3037         goto receivers.
3038         (tree_rest_of_compilation): Register hooks and initialize bitmap
3039         early. Do not set after_inlining flag.
3040
3041 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3042
3043         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3044         HOST_BITS_PER_LONGLONG
3045
3046 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3047
3048         PR c/19977
3049         * c-typeck.c (store_init_value): Don't emit pedantic overflow
3050         warning for non-static initializers.
3051
3052 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
3053
3054         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3055         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3056         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3057         Always use set_unique_reg_note to add REG_EQUAL notes.
3058
3059 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3060
3061         Revert:
3062         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3063
3064         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3065         * cfgbuild.c (find_basic_blocks): Likewise.
3066         * cfgrtl.c (rtl_create_basic_block): Likewise.
3067         * function.c (temp_slots_at_level): Likewise.
3068         * reg-stack.c (stack_regs_mentioned): Likewise.
3069         * regclass.c (allocate_reg_info): Likewise.
3070         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3071         set_bb_for_stmt, move_block_to_fn): Likewise.
3072         * tree-complex.c (tree_lower_complex): Likewise.
3073         * vec.h (VEC_safe_grow_cleared): New.
3074
3075 2007-01-02  Ian Lance Taylor  <iant@google.com>
3076
3077         * c-common.c (c_common_truthvalue_conversion): When warning about
3078         using an assignment as a truth value, set TREE_NO_WARNING.
3079
3080 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3081
3082         PR middle-end/7651
3083         * c.opt (Wold-style-declaration): New.
3084         * doc/invoke.texi (C-only Warning Options): New.
3085         (Wold-style-declaration): Document it.
3086         (Wextra): Enabled by -Wextra.
3087         * c-opts.c (c_common_post_options): Enabled by -Wextra.
3088         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3089         -Wold-style-declaration.
3090
3091 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3092
3093         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3094         * cfgbuild.c (find_basic_blocks): Likewise.
3095         * cfgrtl.c (rtl_create_basic_block): Likewise.
3096         * function.c (temp_slots_at_level): Likewise.
3097         * reg-stack.c (stack_regs_mentioned): Likewise.
3098         * regclass.c (allocate_reg_info): Likewise.
3099         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3100         set_bb_for_stmt, move_block_to_fn): Likewise.
3101         * tree-complex.c (tree_lower_complex): Likewise.
3102         * vec.h (VEC_safe_grow_cleared): New.
3103
3104 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
3105
3106         * c-common.c(c_common_nodes_and_builtins): Since variants of
3107         void_type_node get built before it is given a name, we need to
3108         give those variants the name, too.
3109         (complete_array_type): We need to work with the canonical main
3110         type of the array, from which we will build the qualified version.
3111         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3112         * print-tree.c (print_node): Display canonical type information
3113         for each type.
3114         * stor-layout.c (layout_type): When we don't know the
3115         alignment of a type for which we're building an array, we end up
3116         guessing wrong, so make the type require structural equality.
3117         * tree.c (make_node_stat): When we build a new type, it is its
3118         own canonical type.
3119         (build_type_attribute_qual_variant): When building an attribute
3120         variant, its canonical type is the non-attribute variant. However,
3121         if the attributes are target-dependent and they differ, we need to
3122         use structural equality checks for this type.
3123         (build_qualified_type): A qualified type is not equivalent to its
3124         unqualified variant; set the canonical type appropriately.
3125         (build_distinct_type_copy): When building a distinct type from
3126         another type, the new type is its own canonical type.
3127         (build_variant_type_copy): When building a new type variant, we
3128         assume that it is equivalent to the original type.
3129         (build_pointer_type_for_mode): When building a pointer type, also
3130         build a canonical type pointer.
3131         (build_reference_type_for_mode): When building a reference type,
3132         also build a canonical type reference.
3133         (build_index_type): When we can't hash an index type (e.g.,
3134         because its maximum value is negative), the index type requires
3135         structural equality tests.
3136         (build_array_type): Build the canonical form of an array type.
3137         (build_function_type): Function types require structural equality,
3138         because they contain default arguments, attributes, etc.
3139         (build_method_type_directly): Ditto for method types.
3140         (build_offset_type): Build the canonical offset type.
3141         (build_complex_type): Build the canonical vector type.
3142         (make_vector_type): Build the canonical vector type.
3143         * tree.h (TYPE_CANONICAL): New.
3144         (TYPE_STRUCTURAL_EQUALITY_P): New.
3145         (SET_TYPE_STRUCTURAL_EQUALITY): New.
3146         (struct tree_type): Added "canonical" field.
3147         * params.h (VERIFY_CANONICAL_TYPES): New.
3148         * doc/c-tree.texi (TYPE_CANONICAL): Document.
3149         (TYPE_STRUCTURAL_EQUALITY_P): Document.
3150         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3151         * doc/invoke.texi (verify-canonical-types): Document --param
3152         parameter for verifying canonical types.
3153
3154 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3155
3156         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3157         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3158         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3159         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3160         tm_file.
3161         * config/rs6000/e500.h: New.
3162         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3163         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3164         Remove.
3165         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3166         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3167         TARGET_E500_DOUBLE): Remove.
3168         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3169         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3170         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3171         * config/rs6000/rs6000.c (rs6000_override_options): Use
3172         CHECK_E500_OPTIONS.
3173
3174 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3175
3176         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3177         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3178         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3179         rs6000_emit_cmove): Don't check TARGET_E500.
3180         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3181         sordered): Don't check TARGET_E500.
3182
3183 2007-01-01  Eric Christopher  <echristo@apple.com>
3184
3185         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3186         HARD_REGISTER_NUM_P.
3187
3188 2007-01-01  Roger Sayle  <roger@eyesopen.com>
3189
3190         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3191         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3192         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3193         eq/ne Y".
3194
3195 2007-01-01  Mike Stump  <mrs@apple.com>
3196
3197         * configure.ac: Remove support for building with Apple's gcc-3.1.
3198
3199 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3200
3201         PR middle-end/30311
3202         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3203         REG is a hard register.  Do not modify REG before calling
3204         subreg_nregs.
3205         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3206         multiple of the size of YMODE for certain lowpart cases.
3207
3208 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3209
3210         PR middle-end/30253
3211         * gimplify (voidify_wrapper_expr): Update for
3212         GIMPLE_MODIFY_STMT.
3213
3214 2007-01-01  Andreas Schwab  <schwab@suse.de>
3215
3216         PR target/29166
3217         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3218         save of BR0 in extra_spill_size instead of spill_size.
3219         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3220         area.
3221         (ia64_expand_epilogue): Restore BR0 from its new location.
3222
3223 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3224
3225         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3226         Use a temporary variable if the left hand side is not a gimple
3227         register.
3228
3229 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3230
3231         * gimplify.c (gimplify_return_expr): Make the temporary variable
3232         for the return expression, a gimple register variable.
3233
3234 2007-01-01  Jan Hubicka  <jh@suse.cz>
3235
3236         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3237         INSN_LIST.
3238
3239 2007-01-01  Mike Stump  <mrs@apple.com>
3240
3241         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3242         .literal16.
3243         * config/darwin.c (machopic_select_rtx_section): Use
3244         HAVE_GAS_LITERAL16.
3245         (darwin_mergeable_constant_section): Likewise.
3246         * configure: Regenerate.
3247         * config.in: Regenerate.
3248
3249 2007-01-01  Jan Hubicka  <jh@suse.cz>
3250             Andrew Pinski  <pinskia@gmail.com>
3251
3252         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3253         before starting IPA passes.