OSDN Git Service

cc300099d60c81b27eedb05dc3d90471ac0c90dd
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
2
3         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
4         to handle optabs that operate on floating point input argument and
5         output to integer output.
6         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
7         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
8         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
9         expand_builtin_interclass_mathfn ().
10         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
11         (ilogbsi2): Remove.
12         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
13         ilogbl built-in functions as x87 intrinsics.
14
15 2007-01-30  Richard Guenther  <rguenther@suse.de>
16
17         PR middle-end/27657
18         * dwarf2out.c (reference_to_unused): Query varpool if the
19         variable was output.
20
21 2007-01-30  Richard Guenther  <rguenther@suse.de>
22
23         PR middle-end/30313
24         * passes.c (execute_one_pass): Reset in_gimple_form to not
25         confuse non-unit-at-a-time mode.
26
27 2007-01-29  Roger Sayle  <roger@eyesopen.com>
28             Richard Guenther  <rguenther@suse.de>
29
30         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
31         avoid undefined behaviour on overflow.  Use force_fit_type_double
32         to construct the constant with the specified TREE_OVERFLOW.
33
34 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
35
36         * config/dfp-bit.c: Add parameterized support for fp exceptions.
37         * config/dfp-bit.h: Ditto.
38
39 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
40
41         * c-decl.c (pop_scope): Replace warnings with call to
42         warn_for_unused_label.
43         * c-common.h (warn_for_unused_label): Declare.
44         * c-common.c (warn_for_unused_label): Define.
45
46 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
47
48         * tree-optimize.c (update_inlined_to_pointers): Delete unused
49         function.
50
51 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
52
53         * Makefile.in (USER_H): Remove decfloat.h.
54         * ginclude/decfloat.h: Delete, moving contents to ...
55         * ginclude/float.h: Add support for decimal floating point,
56         guarded by __STDC_WANT_DEC_FP__.
57
58 2007-01-29  Mike Stump  <mrs@apple.com>
59
60         * doc/gccint.texi (Top): Rename Loop Representation to Loop
61         Analysis and Representation to resolve case insensitive conflict.
62         * doc/loop.texi (Loop Analysis and Representation): Likewise.
63
64 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
65
66         * tree.h (struct tree_memory_tag): Add aliases member.
67         (MTAG_ALIASES): New macro.
68         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
69         (add_may_alias): Remove pointer-set. Update for may_aliases being
70         a bitmap. 
71         (mark_aliases_call_clobbered): Update for may_aliases being a
72         bitmap.
73         (compute_tag_properties): Ditto.
74         (create_partition_for): Ditto.
75         (compute_memory_partitions): Ditto.
76         (dump_may_aliases_for): Ditto.
77         (is_aliased_with): Ditto.
78         (add_may_alias_for_new_tag): Ditto.
79         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
80         (compute_is_aliased): New function.
81         (compute_may_aliases): Call compute_is_aliased).
82         (init_alias_info): Initialize alias_bitmap_obstack.
83         (union_alias_set_into): New function.
84         (compute_flow_sensitive_aliasing): Use union_aliases_into.
85         (have_common_aliases_p): Rewrite to take two bitmaps and use
86         intersection.
87         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
88         Update for bitmaps.
89         (finalize_ref_all_pointers): Update for add_may_alias changes.
90         (new_type_alias): Ditto.
91         * tree-flow-inline.h (may_aliases): Return a bitmap.
92         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
93         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
94         may_aliases being a bitmap.
95         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
96         may_aliases now returns a bitmap.
97         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
98         being a bitmap.
99         * tree-ssa-operands.c (add_virtual_operand): Update for
100         may_aliases being a bitmap.
101         
102 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
103
104         PR tree-optimization/30630
105         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
106         variable as changed.
107
108 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
109
110         PR c++/28266
111         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
112         expanded only once even if an error occurs.
113
114 2007-01-29  Ben Elliston  <bje@au.ibm.com>
115
116         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
117         (__gcov_average_profiler): Likewise.
118         (__gcov_ior_profiler): Likewise.
119         (__gcov_merge_ior): Likewise.
120
121 2007-01-28  Jan Hubicka  <jh@suse.cz>
122
123         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
124         * value-prof.c (stringop_block_profile): Likewise.
125
126 2007-01-28  Jan Hubicka  <jh@suse.cz>
127
128         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
129         variant handling histograms; add wrapper.
130         (clear_storage_via_libcall): Export.
131         (emit_block_move_hints): Break out from ...; add histograms.
132         (emit_block_move): ... this one.
133         (clear_storage_hints): Break out from ...; add histograms.
134         (clear_storage): ... this one.
135         (set_storage_via_memset): Handle histogram.
136         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
137         (clear_storage_hints, clear_storage_via_libcall): Declare.
138         (set_storage_via_setmem): Update prototype.
139         * doc/md.texi (movmem, setmem): Document new arguments.
140
141         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
142         new histograms.
143         (stringop_block_profile): New global function.
144         (tree_stringops_values_to_profile): Profile block size and alignment.
145         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
146         HIST_TYPE_IOR.
147         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
148         (stringop_block_profile): Declare.
149         * builtins.c: Include value-prof.h.
150         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
151         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
152         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
153         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
154         * profile.c (instrument_values): Add new counters.
155         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
156         calls.
157         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
158         (tree_init_edge_profiler): Build new profilers.
159         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
160         (pass_tree_profile): Add dump.
161         (tree_profile_hooks): Update.
162         * Makefile.in (LIBGCOV): Add new constants.
163         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
164         __gcov_ior_profiler): New.
165         * i386.md (movmem/setmem expanders): Add new optional arguments.
166
167 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
168
169         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
170
171 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
172
173         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
174         (tree_unswitch_single_loop): Pass eni_size_weights to
175         tree_num_loop_insns.
176         * tree-ssa-loop-manip.c: Include tree-inline.h.
177         (can_unroll_loop_p): Pass eni_size_weights to
178         tree_num_loop_insns.
179         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
180         Pass eni_size_weights to estimate_num_insns.
181         * tree.h (init_inline_once): Export.
182         * toplev.c (backend_init): Call init_inline_once.
183         * cgraphunit.c (cgraph_process_new_functions,
184         cgraph_analyze_function): Pass eni_inlining_weights to
185         estimate_num_insns.
186         * ipa-inline.c (compute_inline_parameters): Ditto.
187         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
188         to estimate_num_insns.
189         (try_unroll_loop_completely): Pass eni_size_weights to
190         tree_num_loop_insns.
191         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
192         ot estimate_num_insns.
193         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
194         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
195         * tree-inline.c (eni_inlining_weights, eni_size_weights,
196         eni_time_weights): New variables.
197         (init_inline_once): Initialize them.
198         (struct eni_data): Mew.
199         (estimate_num_insns_1, estimate_num_insns): Use weights.
200         * tree-inline.h (struct eni_weights_d): New.
201         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
202         (estimate_num_insns): Declaration changed.
203         * cfgloop.h (tree_num_loop_insns): Declaration changed.
204         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
205         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
206
207 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
208
209         * tree-data-ref.c (conflict_fn): Assert that the number of affine
210         relations in the conflict function is valid.
211
212 2007-01-27  Ian Lance Taylor  <iant@google.com>
213
214         * common.opt: Add fstrict-overflow.
215         * opts.c (decode_options): Set flag_strict_overflow if -O2.
216         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
217         (TYPE_OVERFLOW_UNDEFINED): Define.
218         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
219         Replace all uses.
220         * tree.h (TYPE_TRAP_SIGNED): Don't define.
221         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
222         (fold_negate_expr): Likewise.
223         (make_range): Likewise.
224         (extract_muldiv_1): Likewise.
225         (maybe_canonicalize_comparison): Likewise.
226         (fold_comparison): Likewise.
227         (fold_binary): Likewise.
228         (tree_expr_nonnegative_p): Likewise.
229         (tree_expr_nonzero_p): Likewise.
230         * tree-vrp.c (compare_values): Likewise.
231         (extract_range_from_binary_expr): Likewise.
232         (extract_range_from_unary_expr): Likewise.
233         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
234         Likewise.
235         (nowrap_type_p): Likewise.
236         * tree-scalar-evolution.c (simple_iv): Likewise.
237         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
238         (build_range_check): Likewise.
239         (extract_muldiv_1): Likewise.
240         (fold_comparison): Likewise.
241         * tree-vrp.c (vrp_int_const_binop): Likewise.
242         (extract_range_from_unary_expr): Likewise.
243         * convert.c (convert_to_integer): Likewise.
244         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
245         (fold_comparison): Likewise.
246         (fold_binary): Likewise.
247         * optabs.c (optab_for_tree_code): Likewise.
248         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
249         * simplify-rtx.c (simplify_const_relational_operation): Check
250         flag_strict_overflow and flag_trapv.
251         (simplify_const_relational_operation): Likewise.
252         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
253         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
254         -fstrict-overflow.
255
256 2007-01-27  Roger Sayle  <roger@eyesopen.com>
257
258         * tree.c (tree_fold_gcd): Delete.
259         * tree.h (tree_fold_gcd): Remove prototype.
260         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
261         test whether one constant integer is a multiple of another.  Instead
262         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
263         * fold-const.c (multiple_of_p):  We've determined both TOP and
264         BOTTOM are integer constants so we can call int_const_binop directly
265         instead of the more generic const_binop.
266
267 2007-01-27  Roger Sayle  <roger@eyesopen.com>
268
269         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
270         1*X check that the constant hasn't overflowed, to preserve the
271         TREE_OVERFLOW bit.
272         (round_up): Provide an efficient implementation when rouding-up an
273         INTEGER_CST to a power-of-two.
274
275 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
276
277         * doc/sourcebuild.texi: Add comma for clarity.
278         * doc/extend.texi: Fix some typos.
279         * doc/passes.texi: Likewise.
280         * doc/cppinternals.texi: Likewise.
281         * doc/c-tree.texi: Likewise.
282         * doc/tree-ssa.texi: Likewise.
283         * doc/install.texi: Likewise.
284
285 2007-01-27  Jan Hubicka  <jh@suse.cz>
286
287         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
288         (pass_sra): Do not require alias information.
289         * passes.c (init_optimization_passes): Add SRA
290
291 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
292
293         * tracer.c (rest_of_handle_tracer): We already cleaned
294         up the CFG in tracer() so don't do it here again.
295         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
296         cleanup_cfg here, either.  And don't call renumber_insns.
297
298         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
299         and BB_HEAD are in the insn stream here.  Instead make sure
300         that BB_INSN is valid on all insns.  Also, do check here that
301         there are no pending branch predictions...
302         (rtl_verify_flow_info): ...instead of doing it here.  Checks
303         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
304         here.
305
306 2007-01-26  Roger Sayle  <roger@eyesopen.com>
307
308         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
309         function to simplify/factorize operand order canonicalization.
310         (ix86_fixup_binary_operands): Reorganize using the above function.
311         (ix86_binary_operator_ok): Likewise.
312
313 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
314
315         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
316         file scope from optimize_attrs.
317         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
318         up search.
319         (optimize_attrs): Clear insn_code_values after freeing it.
320
321 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
322
323         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
324         a parts.base assignment.
325
326 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
327
328         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
329         (compute_subscript_distance, initialize_data_dependence_relation,
330         finalize_ddr_dependent, analyze_ziv_subscript,
331         analyze_siv_subscript_cst_affine,
332         compute_overlap_steps_for_affine_univar,
333         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
334         analyze_siv_subscript, analyze_miv_subscript,
335         analyze_overlapping_iterations, subscript_dependence_tester_1,
336         compute_self_dependence, free_dependence_relation): Work
337         with affine_fn instead of chrecs.
338         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
339         common_affine_function, affine_function_base,
340         affine_function_constant_p, affine_fn_op, affine_fn_plus,
341         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
342         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
343         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
344         (all_chrecs_equal_p): Removed.
345         * tree-data-ref.h (affine_fn, conflict_function): New types.
346         (struct subscript): Change type of conflicting_iterations_in_a
347         and conflicting_iterations_in_b.
348
349 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
350
351         PR other/30182
352         * config/pa/pa.h (TARGET_HPUX_11): New.
353         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
354         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
355
356 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
357         Richard Guenther  <rguenther@suse.de>
358
359         * tree-ssa-structalias.c (solve_graph): Handle case
360         we merged the variable to another.
361
362 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
363
364         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
365         functions.
366
367         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
368         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
369
370 2007-01-25  DJ Delorie  <dj@redhat.com>
371
372         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
373         allow changes to modes which don't fit in those registers.
374
375         * reload1.c (choose_reload_regs): Check for invalid subregs before
376         computing their locations, not after.
377
378 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
379
380         PR 25127
381         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
382         save Altivec registers in an eh_return function.
383         (compute_vrsave_mask): Likewise.
384         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
385         for saving Altivec registers.
386         (rs6000_emit_prologue): Don't allocate stack twice in
387         eh_return function.  Correct expected value of altivec_save_offset
388         when using save_world.  Describe save of R0 to stack when using
389         save_world.  Describe stack pointer adjustment when using
390         save_world.  Remove duplicated eh_return parameter register saving.
391         Update sp_offset variable after save_world.
392         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
393         darwin-world.asm.
394         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
395         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
396         implies Altivec.
397
398 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
399
400         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
401
402 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
403
404         PR other/30182
405         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
406         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
407
408 2007-01-25  Richard Guenther  <rguenther@suse.de>
409
410         * doc/invoke.texi (-Wcoverage-mismatch): Document.
411         * common.opt (-Wcoverage-mismatch): New warning option.
412         * coverage.c (get_coverage_counts): Ignore coverage mismatch
413         if -Wcoverage-mismatch is given.
414
415 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
416
417         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA. 
418         (ipcp_driver): Change to static definition. 
419         Add dumping of the ifunctions.
420         (constant_val_insert): Remove unused parameter. Support for SSA.
421         (ipcp_propagate_const): Support for SSA.
422         (ipcp_profile_bb_print): Print only analyzed nodes.
423         (ipcp_replace_map_create): Remove support for Fortran constant 
424         for now.
425         * ipa-prop.c (ipa_method_modify_stmt, 
426         ipa_callsite_compute_param): Support for SSA.
427         * ipa-prop.h (ipcp_driver): Remove declaration.
428         (IS_VALID_TREE_MAP_INDEX): Add define.
429
430 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
431
432         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
433
434 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
435
436         PR middle-end/30447
437         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
438         constant argument to cabs and do it without checking for
439         -funsafe-math-optimizations.
440
441 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
442
443         * c-common.h (RID_FIRST_CXX0X): New.
444         (RID_LAST_CXX0X): New.
445         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
446         by -Wall.
447         * c.opt (Wc++0x-compat): New.
448         * doc/invoke.texi (-Wc++0x-compat): Document.
449
450 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
451
452         * builtins.c (fold_builtin_carg): New.
453         (fold_builtin_1): Use it.
454
455 2007-01-24  Jan Hubicka  <jh@suse.cz>
456
457         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
458
459         * ipa-inline.c (initial_insns, max_insns): Delete.
460         (compute_max_insns): New function.
461         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
462         of insns as base for code growth.
463         (cgraph_decide_inlining): Make initial_insns local; do not compute
464         max_insns.
465         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
466         * doc/invoke.texi (inline-unit-growth): Update docs.
467
468 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
469
470         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
471         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
472         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
473         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
474         for CPUs that have PTA_CX16 set.
475
476         PR middle-end/27416
477         * gimplify.c (omp_check_private): New function.
478         (gimplify_scan_omp_clauses): Use it for
479         firstprivate/lastprivate/reduction.
480
481         PR middle-end/30494
482         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
483         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
484
485         PR middle-end/30421
486         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
487         If lower_lastprivate_clauses emits some statements, append them
488         to dlist rather than body_p and to body_p append an initializer.
489         (lower_omp_for): Adjust caller.
490
491 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
492
493         * target.h (globalize_decl_name): New.
494         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
495         * output.h (default_globalize_decl_name): New.
496         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
497         globalize_label.
498         (globalize_decl): Ditto.
499         (default_globalize_decl_name): New.
500         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
501         (ia64_handle_version_id_attribute): New.
502         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
503         (ia64_asm_output_external): Use globalize_decl_name instead
504         of globalize_label.
505         * doc/extend.texi (version_id): New pragma.
506         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
507
508 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
509
510         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with 
511         _uleb128_t and _Unwind_SWord with _sleb128_t.
512         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
513         uw_frame_state_for, uw_update_context_1): Likewise.
514         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
515         * unwind-pe.h (read_uleb128, read_sleb128,
516         read_encoded_value_with_base): Likewise.
517         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
518
519 2007-01-24  Richard Guenther  <rguenther@suse.de>
520
521         * builtins.c (expand_builtin_cexpi): Get the fndecl
522         for cexp in the correct way.
523
524 2007-01-24  Jan Hubicka  <jh@suse.cz>
525
526         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
527
528 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
529
530         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
531         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
532         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
533         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
534         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
535         __SIZEOF_PTRDIFF_T__.
536         * doc/cpp.texi: Documentation for the new builtin macros added.
537
538 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
539
540         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
541         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
542         instead of operands[2] to avoid access past the end of array.
543
544 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
545
546         * reload1.c (emit_reload_insns): Pass the reload register
547         for a non-spill output reload through forget_old_reloads_1.
548
549 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
550
551         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
552         _SOFT_DOUBLE if doubles use software floating-point.
553         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
554         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
555         * config/rs6000/darwin-ldouble.c: Also compile functions for
556         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
557         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
558         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
559         commented-out long double override.
560         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
561         * config/rs6000/eabispe.h: Likewise.
562         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
563         long double for non-SPE.
564         (rs6000_handle_option): Likewise.
565         (invalid_e500_subreg): Disallow more subregs involding DImode,
566         DFmode, TImode or TFmode.
567         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
568         E500 double.
569         (legitimate_lo_sum_address_p): Also check for TFmode for E500
570         double.
571         (rs6000_legitimize_address): Also handle TFmode for E500 double.
572         (rs6000_legitimize_reload_address): Also handle TFmode for E500
573         double.
574         (rs6000_legitimate_address): Also check for TFmode for E500
575         double.
576         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
577         (spe_build_register_parallel): Handle TFmode and TCmode.
578         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
579         double.
580         (function_arg): Handle TFmode and TCmode for E500 double.
581         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
582         in general.
583         (print_operand): Handle TFmode and TImode for %y.
584         (rs6000_generate_compare): Handle TFmode comparisons for E500
585         double.
586         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
587         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
588         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
589         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
590         TFmode for E500 double.
591         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
592         (floatsidf2): Enable for E500 double.
593         (movtf_softfloat): Use rs6000_nonimmediate_operand.
594         (extenddftf2): Change to extenddftf2_fprs.
595         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
596         depending on TARGET_E500_DOUBLE.
597         (extendsftf2): Enable for E500 double.
598         (trunctfdf2): Enable for E500 double.
599         (trunctfsf2): Change to trunctfsf2_fprs.
600         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
601         depending on TARGET_E500_DOUBLE.
602         (floatsitf2): Enable for E500 double.
603         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
604         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
605         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
606         (negtf2): Change to negtf2_internal.
607         (negtf2): New expander.
608         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
609         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
610         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
611         (movdi_internal32): Use rs6000_nonimmediate_operand.
612         (unnamed splitter): Likewise.
613         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
614         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
615         (SPE64TF, DITI): New mode macros.
616         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
617         modes.
618         (frob_tf_ti): New.
619         (frob_<mode>_di_2): New.
620         (frob_tf_di_8_2): New.
621         (frob_di_df): Change to frob_di_<mode>; allow more modes.
622         (frob_ti_tf): New.
623         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
624         more modes.
625         (frob_ti_<mode>_8_2): New.
626         (frob_ti_tf_2): New.
627         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
628         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
629         TFmode.
630         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
631         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
632         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
633         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
634         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
635         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
636         tsttflt_gp): New.
637
638 2007-01-23  Ian Lance Taylor  <iant@google.com>
639
640         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
641         insn-*.o first.
642         (OBJS-archive): Reformat, alphabetize.
643         (OBJS): Change out_object_file to OBJS-md.
644
645 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
646
647         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
648         JUMP_P and CALL_P predicates where applicable.
649         * config/i386/i386.c: Ditto.
650         * config/i386/i386.md: Ditto.
651         * config/i386/mmx.md: Ditto.
652         * config/i386/predicates.md: Ditto.
653
654 2007-01-22  Andreas Schwab  <schwab@suse.de>
655
656         * config/m68k/m68k.h: Fix comment.
657
658 2007-01-22  Jan Hubicka  <jh@suse.cz>
659
660         * passes.c (init_optimization_passes): Do not rerun
661         pass_early_warn_uninitialized.
662
663 2007-01-22  Richard Guenther  <rguenther@suse.de>
664
665         PR tree-optimization/30038
666         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
667         function.
668         (execute_cse_sincos_1): Likewise.
669         (execute_cse_sincos): Likewise.
670         (gate_cse_sincos): Likewise.
671         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
672         the cexpi() canonicalization of sincos().
673         * tree-pass.h (pass_cse_sincos): Declare.
674         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
675
676 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
677
678         PR rtl-optimization/29329
679         * combine.c (replaced_rhs_insn): Rename to i2mod.
680         (replaced_rhs_value): Rename to i2mod_new_rhs.
681         (i2mod_old_rhs): New global variable.
682         (combine_instructions): Adjust for above change.  Save a copy of
683         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
684         note are substituted in the second instruction.
685         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
686         ditch the note if it pertains to the second eliminated register
687         and this register is mentioned in i2mod_old_rhs.
688
689         Revert:
690         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
691
692         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
693         the insn to which the note was originally attached.
694
695 2007-01-21  Jan Hubicka  <jh@suse.cz>
696
697         * ipa-inline.c (inlining_mode): Comment, move up.
698         (cgraph_decide_inlining_incrementally): Do not perform inlining
699         itself; fix handling of flattening of self recursive functions.
700         (cgraph_find_cycles): Remove.
701         (cgraph_flatten_node): Remove.
702         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
703         (try_inline): New function.
704         (cgraph_early_inlining): Update call of
705         cgraph_decide_inlining_incrementally.  Apply inlining here.
706         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
707
708 2007-01-21  Dirk Mueller  <dmueller@suse.de>
709
710         PR bootstrap/30511
711         * tree-vrp.c (check_array_bounds): do not warn
712         about ADDR_EXPR's of ARRAY_REF's which are immediately
713         used in binary expressions.
714
715 2007-01-21  Ira Rosen  <irar@il.ibm.com>
716
717         * tree-vectorizer.h (struct _stmt_vec_info): Add new field 
718         read_write_dep and macros for its access.
719         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
720         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove 
721         argument, call vect_check_interleaving for every independent pair of
722         data-refs. Mark loads that access the same memory location as a store
723         in the loop.
724         (vect_check_dependences): Remove.
725         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences 
726         call, fix the call to vect_analyze_data_ref_dependence.
727         (vect_analyze_data_ref_access): For statements that access the same 
728         data-ref, check that they are not stores; for loads, check that there
729         is no store that access the same location.
730
731 2007-01-20  Roger Sayle  <roger@eyesopen.com>
732             Joseph Myers  <joseph@codesourcery.com>
733
734         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
735         (-fdump-rtl-stack): Likewise.
736         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
737         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
738         (-mpt-fixed): Use @minus{} for minus sign.
739         (Using Precompiled Headers): Fix under/overfull hbox.
740
741 2007-01-20  Jan Hubicka  <jh@suse.cz>
742
743         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
744         place within annotation.
745
746 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
747
748         * pa.c (output_move_double): Change array size of xoperands to 4.
749
750 2007-01-20  Jan Hubicka  <jh@suse.cz>
751
752         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
753         as temporary.
754
755 2007-01-19  Ian Lance Taylor  <iant@google.com>
756
757         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
758         constant to unsigned HOST_WIDE_INT before negating.
759
760 2007-01-19  Ian Lance Taylor  <iant@google.com>
761
762         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
763         to unsigned.
764         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
765         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
766         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
767         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
768         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
769         (realloc_vdef, realloc_vuse): Change second parameter to
770         unsigned.
771         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
772         fields to unsigned.
773         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
774         unsigned. Change x and lim locals to unsigned.
775         (realloc_vdef, realloc_vuse): Change num_elem parameter to
776         unsigned.
777         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
778         (copy_virtual_operands): Change i and n locals to unsigned.
779
780 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
781
782         PR middle-end/29335
783         * builtins.c (fold_builtin_1): Handle builtin fdim.
784
785 2007-01-20  Jan Hubicka  <jh@suse.cz>
786
787         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
788         * tree-ssa-structalias.c (compute_points_to_sets): Do call
789         init_alias_heapvars.
790         (init_alias_heapvars): Initialize only when not already initialized.
791         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
792
793 2007-01-19  Roger Sayle  <roger@eyesopen.com>
794
795         * common.opt (fsigned-zeros): New command line option.
796         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
797         of flag_unsafe_math_optimizations.
798         * opts.c (set_fast_math_flags): The -ffast-math command line option
799         implies -fno-signed-zeros.
800         (fast_math_flags_set_p): Likewise.
801
802         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
803         the documentation of -ffast-math appropriately.  Wrap long lines.
804
805 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
806
807         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
808         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
809
810 2007-01-19  Tomas Bily  <tbily@suse.cz>
811
812         * cgraphunit.c (cgraph_finalize_function): Updating of pid
813         * tree-profile.c:
814         (tree_init_ic_make_global_vars): New function
815         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
816         (tree_gen_ic_profiler): New function
817         (tree_gen_ic_func_profiler): New function
818         (tree_profiling): Added calling of tree_gen_ic_func_profiler
819         (tree_profile_hooks): Added hook for indirec/virtual calls
820         * value-prof.c (tree_find_values_to_profile): New case for
821         indirect calls
822         (tree_values_to_profile): Call for determining indirect/virtual
823         counters
824         (tree_indirect_call_to_profile): New function
825         (tree_ic_transform): New function
826         (tree_ic): New function
827         (find_func_by_pid): New function
828         (init_pid_map): New function
829         (tree_value_profile_transformations): Added check for
830         indirect/virtual call transformation
831         * value-prof.h (enum hist_type): New counter type for
832         indirect/virtual calls
833         (profile_hooks): Added new hook for profiling indirect/virtual
834         calls
835         * profile.c (instrument_values): New case for indirect/virtual
836         call added
837         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
838         (GCOV_COUNTER_V_INDIR): New counter type
839         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
840         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
841         call added
842         * cgraph.c: Definition of cgraph_max_pid
843         (cgraph_create_node): Default init of pid attribute
844         * cgraph.h: Declaration of cgraph_max_pid
845         (struct cgraph_node): Added pid attribute
846         * libgcov.c (__gcov_indirect_call_profiler): New function
847         (__gcov_one_value_profiler_body): New function
848         (__gcov_one_value_profiler): Body was moved to
849         __gcov_one_value_profiler_body and calls it
850
851 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
852         
853         * doc/gty.texi (Options): Document the mark_hook option to GTY.
854         * gengtype.c (write_types_data, write_func_for_structure,
855         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
856         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
857         (walk_type, write_func_for_structure): Generate the mark_hook if
858         needed.
859
860 2007-01-19  Jan Hubicka  <jh@suse.cz>
861
862         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
863         'early' argument take inlining mode argument specifying whether to
864         inline for size/speeed or all functions; add support for flattening;
865         improve dumpting.
866         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
867
868 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
869
870         PR c++/17947
871         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
872         the name of the declared identifier.
873                 
874 2007-01-19  Dirk Mueller  <dmueller@suse.de>
875
876         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
877         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
878         duplicate evaluation.
879
880 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
881
882         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
883         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
884         to XFMode, use acosxf2 and truncate result to requested mode.
885         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
886         (asin<mode>2): Similarly, with asin expanders.
887         (*fscalexf4_i387): Rename from *fscalexf4.
888         (expNcorexf3): New expander.
889         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
890         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
891         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
892         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
893         patterns for SSE math.
894         (exp10<mode>2): Similarly, with exp10 expanders.
895         (exp2<mode>2): Similarly, with exp2 expanders.
896         (expm1<mode>2): Similarly, with expm1 expanders.
897         (ldexp<mode>3): Similarly, with ldexp expanders.
898         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
899         Use gen_truncxf<mode>2_i387_noop to truncate result.
900
901 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
902
903         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
904         for ColdFire targets.
905
906 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
907             Richard Sandiford  <richard@codesourcery.com>
908
909         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
910         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
911         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
912         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
913         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
914         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
915
916 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
917
918         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
919         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
920         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
921         (*adddi_dilshr32_cf): New pattern.
922         (adddi3, subdi3): Remove first alternatives.
923
924 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
925
926         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
927         implemented using move.l, do not use its cc result for floating-point
928         comparisons.
929
930 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
931
932         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
933         from accepting 's' constraints if flag_pic.
934
935 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
936
937         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
938         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
939         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
940         condition from TARGET_68881 to TARGET_HARD_FLOAT.
941
942 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
943
944         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
945         ColdFire alternatives.
946         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
947         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
948         architectures that support ff1 instruction.
949
950 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
951             Julian Brown  <julian@codesourcery.com>
952
953         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
954         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
955         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
956         Add commentary.
957
958 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
959
960         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
961         and return a bool.
962         (output_move_const_into_data_reg, output_move_simode_const): Delete.
963         * config/m68k/m68k.c (const_method, const_int_cost): Take a
964         HOST_WIDE_INT instead of an rtx.
965         (m68k_rtx_costs): Update call accordingly.
966         (output_move_const_into_data_reg): Likewise.  Fix formatting.
967         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
968         Return a bool.
969         (output_move_simode_const): Update calls after above changes.
970         Rework to use automatic variables and predicates like MEM_P.
971         * config/m68k/m68k.md (pushexthisi_const): Update call to
972         valid_mov3q_const.
973
974 2007-01-19  Dirk Mueller  <dmueller@suse.de>
975
976         * tree-ssa-alias.c (perform_var_substitution): Fix typo
977         in dump_flags test.
978
979 2007-01-19  Richard Guenther  <rguenther@suse.de>
980
981         * builtins.c (expand_builtin_cexpi): Fall back to expanding
982         via cexp in case sincos is not available.
983
984 2007-01-19  Richard Guenther  <rguenther@suse.de>
985
986         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
987         * defaults.h (TARGET_HAS_SINCOS): Default to off.
988         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
989         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
990         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
991         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
992         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
993         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
994
995 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
996
997         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
998         New insn patterns.
999         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1000         (atan2xf3): Directly generate RTL pattern.
1001         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1002         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1003         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1004         disable patterns for SSE math.
1005         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1006         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1007         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1008         disable patterns for SSE math.
1009
1010 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1011
1012         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1013
1014 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1015
1016         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1017         VIEW_CONVERT_EXPR to the same type as it's operand.
1018
1019 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1020
1021         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1022         configured for long double 128.
1023
1024 2007-01-18  Mike Stump  <mrs@apple.com>
1025
1026         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1027         error.
1028
1029 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1030
1031         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1032         type correct.
1033
1034 2007-01-18  Jan Hubicka  <jh@suse.cz>
1035
1036         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1037         greater than NUM_VOP_FREE_BUCKETS.
1038
1039 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1040
1041         * tree-ssa-structalias.c: Update comments.
1042         (ptabitmap_obstack): Removed.
1043         (pta_obstack): New.
1044         (oldpta_obstack): Ditto.
1045         (stats): Add a few members.
1046         (struct variable_info): Remove node, complex, address_taken, and
1047         indirect_target members. Add oldsolution member.
1048         (new_var_info): Do not initialize removed members.
1049         (constraint_expr_type): Remove INCLUDES.
1050         (constraint_graph): Add size, implicit_preds, rep,
1051         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1052         (FIRST_REF_NODE): New macro.
1053         (LAST_REF_NODE): Ditto.
1054         (FIRST_ADDR_NODE): Ditto.
1055         (find): New function.
1056         (unite): Ditto.
1057         (dump_constraint): Do not handle INCLUDES.
1058         (insert_into_complex): Do not insert duplicate constraints.
1059         (condense_varmap_nodes): Renamed and rewritten into ...
1060         (merge_node_constraints): This. Also fix bug in handling of
1061         offseted copy constraints.
1062         (clear_edges_for_node): No longer need to deal with preds at all,
1063         or removing associated preds/succs. 
1064         (merge_graph_nodes): Deal with indirect_cycles.
1065         Don't deal with predecessors.
1066         (add_implicit_graph_edge): New function.
1067         (add_pred_graph_edge): Ditto.
1068         (add_graph_edge): Don't deal with predecessors.
1069         (build_constraint_graph): Removed.
1070         (build_pred_graph): New function.
1071         (build_succ_graph): Ditto.
1072         (struct scc_info): Removed in_component. Added roots, dfs, and
1073         node_mapping. Remove visited_index, unification_queue.
1074         (scc_visit): Deal with union-find we do now.
1075         Deal with cycles with REF nodes.
1076         (collapse_nodes): Renamed and rewritten to ...
1077         (unify_nodes): This.
1078         (process_unification_queue): Removed.
1079         (topo_visit): Cleanup
1080         (do_da_constraint): Use find.
1081         (do_sd_constraint): Ditto.
1082         (do_ds_constraint): Ditto.
1083         (do_complex_constraint): Ditto.
1084         (init_scc_info): Update for removed and added members.
1085         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1086         (find_indirect_cycles): This.
1087         (equivalence_class): New variable.
1088         (label_visit): New function.
1089         (perform_variable_substitution): Rewritten.
1090         (free_var_substitution_info): New function.
1091         (find_equivalent_node): Ditto.
1092         (move_complex_constraints): Ditto.
1093         (eliminate_indirect_cycles): Ditto.
1094         (solve_graph): Only propagate changed bits.
1095         Use indirect cycle elimination.
1096         Use find.
1097         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1098         (tree_id_eq): Renamed to ...
1099         (tree_vi_eq): This. Update for member change
1100         (insert_id_for_tree): Renamed and rewritten to ...
1101         (insert_vi_for_tree): This.
1102         (lookup_id_for_tree): Renamed and rewritten to ...
1103         (lookup_vi_for_tree): This.
1104         (get_id_for_tree): Renamed and rewritten to ...
1105         (get_vi_for_tree): Ditto.
1106         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1107         (process_constraint): Don't handle INCLUDES.
1108         Remove special ADDRESSOF case.
1109         (find_func_aliases): Rewrite to use vi functions instead of id
1110         ones.
1111         (create_function_info_for): Ditto.
1112         (create_variable_info_for): Ditto.
1113         (intra_create_variable_infos): Ditto.
1114         (merge_smts_into): Ditto.
1115         (find_what_p_points_to): Ditto.
1116         (init_base_vars): Ditto.
1117         (init_alias_vars): Ditto.
1118         (remove_preds_and_fake_succs): New function.
1119         (dump_sa_points_to_info): Dump new stats.
1120         (dump_solution_for_var): Use find.
1121         (set_used_smts): Fix formatting.
1122         (compute_points_to_sets): Updated for new functions.
1123         (ipa_pta_execute): Ditto.
1124
1125 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1126             Richard Sandiford  <richard@codesourcery.com>
1127
1128         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1129         parallels.
1130         * calls.c (expand_call): If the return value is a PARALLEL,
1131         extract its first member.
1132         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1133         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1134         use FUNCTION_EXTRA_EPILOGUE.
1135         (m68k_function_value): Return a PARALLEL if the return value
1136         is of a pointer type.
1137         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1138         (FUNCTION_EXTRA_EPILOGUE): Remove.
1139         * config/m68k/m68k.md (D0_REG): New constant.
1140
1141 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1142
1143         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1144         output a NOP for empty epilogues.
1145
1146 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1147
1148         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1149         before function.  Extend register save check to include all
1150         registers, not just integer ones.
1151
1152 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1153
1154         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1155         (m68k_use_return_insn): ...this.
1156         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1157         * config/m68k/m68k.c (use_return_insn): Rename to...
1158         (m68k_use_return_insn): ...this.
1159         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1160         of USE_RETURN_INSN.
1161
1162 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1163
1164         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1165         nearest even, fix denormal rounding overflow.
1166
1167 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1168
1169         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1170         f%$smove and f%$move.
1171         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1172         (extendsfdf2_cf): Use fdmove instead of f%&move.
1173         (truncdfsf2_cf): Use fsmove instead of f%$smove.
1174         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1175
1176 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1177
1178         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1179         GPR<-GPR moves.
1180
1181 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1182
1183         * real.h (real_format): Add a canonical_nan_lsbs_set field.
1184         (coldfire_single_format): Declare.
1185         (coldfire_double_format): Likewise.
1186         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1187         of qnan_msb_set to determine the lower bits of a canonical
1188         NaN significand.
1189         (encode_ieee_double): Likewise.
1190         (encode_ieee_quad): Likewise.
1191         (ieee_single_format): Initialize canonical_nan_lsbs_set.
1192         (mips_single_format): Likewise.
1193         (ieee_double_format): Likewise.
1194         (mips_double_format): Likewise.
1195         (ieee_extended_motorola_format): Likewise.
1196         (ieee_extended_intel_96_format): Likewise.
1197         (ieee_extended_intel_128_format): Likewise.
1198         (ieee_extended_intel_96_round_53_format): Likewise.
1199         (ibm_extended_format): Likewise.
1200         (mips_extended_format): Likewise.
1201         (ieee_quad_format): Likewise.
1202         (mips_quad_format): Likewise.
1203         (vax_f_format): Likewise.
1204         (vax_d_format): Likewise.
1205         (vax_g_format): Likewise.
1206         (i370_single_format): Likewise.
1207         (i370_double_format): Likewise.
1208         (decimal_single_format): Likewise.
1209         (decimal_double_format): Likewise.
1210         (decimal_quad_format): Likewise.
1211         (c4x_single_format): Likewise.
1212         (c4x_extended_format): Likewise.
1213         (real_internal_format): Likewise.
1214         (coldfire_single_format): New real_format.
1215         (coldfire_double_format): Likewise.
1216         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1217         canonical_nan_lsbs_set.
1218         (pdp11_d_format): Likewise.
1219         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1220         if TARGET_COLDFIRE_CPU.
1221
1222 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1223
1224         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1225         (output_call, m68k_legitimize_call_address): Declare.
1226         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1227         parenthesees.  Add support for a 'W' constraint.
1228         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1229         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1230         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1231         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1232         variables.
1233         (override_options): Initialize them.  Do not set flag_no_function_cse
1234         for TARGET_ID_SHARED_LIBRARY.
1235         (m68k_output_pic_call): Delete.
1236         (m68k_legitimize_call_address): New function.
1237         (print_operand): Remove the %o prefix.  Handle the %p prefix.
1238         (output_call): New function.
1239         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
1240         target address from the GOT if symbolic jumps are not allowed.
1241         * config/m68k/m68k.md (call, general_operand): Do not set
1242         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
1243         Merge separate flag_pic and !flag_pic define_insns into...
1244         (*call, *call_value): ...these new patterns.  Match the address
1245         rather than the containing MEM and require it to be a call_operand.
1246         Use output_call to generate the asm template.
1247         * config/m68k/predicates.md (const_call_operand): New predicate.
1248         (call_operand): Likewise.
1249
1250 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1251
1252         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1253         (STACK_GROWS_DOWNWARD): Define to 1.
1254         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1255         comments, emphasizing that these values are only defaults.
1256         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1257         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1258
1259 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1260
1261         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1262
1263 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1264             Richard Sandiford  <richard@codesourcery.com>
1265
1266         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1267         (FP_REGNO_P): New macros.
1268         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1269         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1270         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1271         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1272
1273 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1274
1275         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1276         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1277         to $tmake_file.
1278         * config/m68k/t-floatlib: New file, extracting common code from...
1279         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1280         * config/m68k/t-uclinux: Here.
1281         * config/m68k/fpgnulib.c: Do not compile extendeed precision
1282         routines on ColdFire targets.
1283
1284 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1285
1286         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1287         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1288         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1289         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1290         explicitly set MOTOROLA to 1.
1291         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1292
1293 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1294             Nathan Sidwell  <nathan@codesourcery.com>
1295
1296         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1297         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1298         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1299         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1300         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1301         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1302         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1303         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1304         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1305         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1306         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1307         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1308         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1309         (DBX_REGISTER_NUMBER): Delete.
1310         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1311         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1312         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1313
1314 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1315
1316         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1317         * config/flat.h: New file.
1318         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1319         OBJECT_FORMAT_FLAT.
1320         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1321         indirect references for -msep-data or -mid-shared-library.
1322         Do not use PC-relative code addresses either.
1323         * config/m68k/m68k.c (override_options): Restrict -fPIC error
1324         to -mpcrel.
1325         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
1326         for shared libraries and crt1.o for executables.  Use crti.o and
1327         crtbegin.o.
1328         (ENDFILE_SPEC): Use crtend.o and crtn.o.
1329         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1330         Do not add -elf2flt or -shared-lib-id options here.
1331         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
1332         Pass -shared-lib-id if -mid-shared-library, taking the library
1333         identifier from -mshared-library-id if given, otherwise
1334         defaulting to 0.
1335         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1336         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1337         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1338         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1339         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1340         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1341         and crtend.o.
1342         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1343         jump sequence for ISA A and ISA A+.
1344         (PICJUMP): Likewise.
1345
1346 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1347
1348         * config.gcc (m68k-*-uclinux*): Base the port on the common
1349         and m68k GNU/Linux files rather than on the generic ELF ones.
1350         * config/m68k/uclinux.h (TARGET_VERSION): Override.
1351         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1352
1353 2007-01-18  Julian Brown  <julian@codesourcery.com>
1354             Richard Sandiford  <richard@codesourcery.com>
1355
1356         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1357         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1358         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1359         before redefining.
1360         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1361         80 unconditionally.
1362         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1363
1364 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1365
1366         * doc/install.texi: Document m68k-uclinuxoldabi.
1367         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1368         * config/m68k/uclinux-oldabi.h: New file, copied from
1369         config/m68k/uclinux.h.
1370
1371 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
1372
1373         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1374         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1375         (m32r_compute_frame_size): Use unsigned for regno.
1376         (m32r_reload_lr): Use gen_frame_mem.
1377         (pop): New.
1378         (m32r_output_function_epilogue): Don't output the function epilogue
1379         textually here.
1380         (m32r_expand_epilogue): New.
1381         (direct_return): Return false if the function has the interrupt
1382         attribute.
1383         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1384         * config/m32r/m32r.md (epilogue): New expander.
1385         (return_lr, return_rte): New insns.
1386         (return): Make it expander.
1387         (return_normal): New expander.
1388
1389 2007-01-18  Josh Conner  <jconner@apple.com>
1390
1391         PR target/30485
1392         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1393         support for UNLE, UNLT, UNGE, and UNGT.
1394
1395 2007-01-18  Jan Hubicka  <jh@suse.cz>
1396
1397         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1398         * tree-tailcall (add_virtual_phis): Likewise.
1399         (optimize_tail_call): Return TODOs.
1400         (execute_tail_calls): Return TODOs.
1401         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1402         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1403         changed.
1404         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1405         Cleanup cfg using TODO.
1406         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1407         * passes.c (execute_function_todo): When cleanup did something, remove
1408         unused locals.
1409         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1410         (make_edges): Don't cleanup_cfg.
1411
1412 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
1413
1414         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1415         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1416         operate on the top of stack.
1417         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1418         register handling of instructions that output to the second
1419         stack slot.
1420         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1421         (move_for_stack_reg): Special-case check for dead destination
1422         stack slot for constant load of 1.0 inside UNSPEC_TAN.
1423
1424         * config/i386/i386.md (UNSPEC_TAN): New constant.
1425         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1426         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1427         to correctly model move of constant 1.0 to top stack slot.
1428         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1429         (unnamed peephole2 pattern): Remove corresponding peephole2
1430         pattern that optimizes tan insn and loading of constant 1.0.
1431         (tanxf2): Use fptanxf4_i387.
1432         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1433         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
1434         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1435         disable patterns for SSE math.
1436         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1437         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1438
1439 2007-01-18  Dirk Mueller  <dmueller@suse.de>
1440             Richard Guenther <rguenther@suse.de>
1441
1442         PR diagnostic/8268
1443         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1444         * common.opt (Warray-bounds): Add new warning option.
1445         * c-opts.c (c_common_handle_option): Define -Warray-bounds
1446         if -Wall is given.
1447         * Makefile.in: make tree-vrp.o depend on toplev.h
1448         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1449         is enabled.
1450         (check_array_refs, check_array_bounds, check_array_ref): New.
1451
1452 2007-01-18  Jan Hubicka  <jh@suse.cz>
1453
1454         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1455         (execute_ssa_ccp): Return flags conditionally.
1456         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1457         changed.
1458         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1459
1460 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
1461
1462         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1463         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1464
1465         * Makefile.c (GTFILES): Add cfglayout.h.
1466         * gengtype.c (open_base_files): Likewise.
1467         * cfglayout.c (cfg_layout_function_footer,
1468         cfg_layout_function_header) Reindent to make gengtype happy.
1469         * cfglayout.h (cfg_layout_function_footer,
1470         cfg_layout_function_header): Add GTY(()) marker.
1471
1472         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1473
1474 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1475
1476         * genautomata.c (write_automata): Include xstrerror output in the
1477         error message if writing the DFA description file fails.
1478
1479 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1480
1481         * config/mips/mips-protos.h (mips_output_external): Make it
1482         return void.
1483         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1484         * config/mips/mips.c (irix_output_external_libcall): Likewise.
1485         (extern_list): Likewise.
1486         (extern_head): Likewise.
1487         (TARGET_ASM_FILE_END): Likewise.
1488         (mips_file_end): Likewise.
1489         (mips_output_external): Rewritten.
1490
1491 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1492
1493         * genpreds.c (write_insn_preds_c): Only write out the function
1494         body for regclass_for_constraint if we have register constraints.
1495
1496 2007-01-17  Tom Tromey  <tromey@redhat.com>
1497
1498         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1499         * doc/install.texi (Testing): Use sourceware.org.
1500         (Binaries): Likewise.
1501         (Specific): Likewise.
1502         * doc/contrib.texi (Contributors): Use sourceware.org.
1503
1504 2007-01-17  Anatoly Sokolov <aesok@post.ru>
1505
1506         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1507         (AVR_ENHANCED): Rename to ...
1508         (AVR_HAVE_MUL): ... new.
1509         (avr_enhanced_p): Rename to ...
1510         (avr_have_mul_p): ... new.
1511         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of 
1512         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1513         * config/avr/avr.c (avr_enhanced_p): Rename to ...
1514         (avr_have_mul_p): ... new.
1515         (base_arch_s): Rename 'enhanced' to 'have_mul'.
1516         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1517         'avr_enhanced_p' and 'enhanced'.
1518         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use 
1519         AVR_HAVE_MUL instead of AVR_ENHANCED.
1520         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1521         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3, 
1522         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL 
1523         instead of AVR_ENHANCED.
1524         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1525         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1526         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1527
1528 2007-01-17  Ian Lance Taylor  <iant@google.com>
1529
1530         * vec.h (VEC_reserve_exact): Define.
1531         (vec_gc_p_reserve_exact): Declare.
1532         (vec_gc_o_reserve_exact): Declare.
1533         (vec_heap_p_reserve_exact): Declare.
1534         (vec_heap_o_reserve_exact): Declare.
1535         (VEC_OP (T,A,reserve_exact)): New static inline function, three
1536         versions.
1537         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1538         negative parameter.
1539         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1540         (VEC_OP (T,A,copy)) [all versions]: Likewise.
1541         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1542         * vec.c (calculate_allocation): Add exact parameter.  Change all
1543         callers.
1544         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1545         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1546         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1547         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1548         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1549         (vec_heap_p_reserve_exact): New function.
1550         (vec_heap_o_reserve_exact): New function.
1551
1552 2007-01-17  Jan Hubicka  <jh@suse.cz>
1553
1554         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1555
1556 2007-01-17  Eric Christopher  <echristo@apple.com>
1557
1558         * config.gcc: Support core2 processor.
1559
1560 2007-01-16  Jan Hubicka  <jh@suse.cz>
1561
1562         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1563         the names, just unlink the chain so we don't crash on dangling pointers
1564         to dead SSA names.
1565
1566 2007-01-16  Jan Hubicka  <jh@suse.cz>
1567
1568         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1569         * tree-pass.h: Reorder to make IPA passes appear toegher.
1570         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1571         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1572         parameters, do not call early inliner.
1573         * ipa-inline.c: Update comments.  Include tree-flow.h
1574         (cgraph_decide_inlining): Do not compute inlining parameters.
1575         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1576         be called with function context set up.
1577         (pass_ipa_inline): Remove unreachable functions before pass.
1578         (cgraph_early_inlining): Simplify assuming to be called from the
1579         PM as local pass.
1580         (pass_early_inline): New pass.
1581         (cgraph_gate_ipa_early_inlining): New gate.
1582         (pass_ipa_early_inline): Turn into simple wrapper.
1583         (compute_inline_parameters): New function.
1584         (gate_inline_passes): New gate.
1585         (pass_inline_parameters): New pass.
1586         (apply_inline): Move here from tree-optimize.c
1587         (pass_apply_inline): New pass.
1588         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1589         transforming.
1590         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1591         doing them by hand.
1592         (tree_function_versioning): Do not allocate dummy struct function.
1593         * tree-inline.h (optimize_inline_calls): Update prototype.
1594         * tree-optimize.c (execute_fixup_cfg): Export.
1595         (pass_fixup_cfg): Remove
1596         (tree_rest_of_compilation): Do not apply inlines.
1597         * tree-flow.h (execute_fixup_cfg): Declare.
1598         * Makefile.in (gt-passes.c): New.
1599         * passes.c: Include gt-passes.h
1600         (init_optimization_passes): New passes.
1601         (nnodes, order): New static vars.
1602         (do_per_function_toporder): New function.
1603         (execute_one_pass): Dump current pass here.
1604         (execute_ipa_pass_list): Don't dump current pass here.
1605
1606 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
1607
1608         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1609         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1610
1611 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
1612
1613         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1614         (strong_alias): Define.
1615         (__gcc_qmul): Provide non-FMA for soft-float.
1616         (__gcc_qdiv): Same.
1617         (__gcc_qneg): New.
1618         (__gcc_qeq): New.
1619         (__gcc_qle): New.
1620         (__gcc_qge): New.
1621         (__gcc_qunord): New.
1622         (__gcc_stoq): New.
1623         (__gcc_dtoq): New.
1624         (__gcc_qtos): New.
1625         (__gcc_qtod): New.
1626         (__gcc_qtoi): New.
1627         (__gcc_qtou): New.
1628         (__gcc_itoq): New.
1629         (__gcc_utoq): New.
1630         (fmsub): New.
1631         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1632         soft-float functions.
1633         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1634         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1635         about long double soft float.
1636
1637 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
1638             Tehila Meyzels  <tehila@il.ibm.com>
1639
1640         * tree-vectorizer.h (is_pattern_stmt_p): New.
1641         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1642         formatting (tabs instead of spaces). Cleanup and clarify setting
1643         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1644         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1645         (vectorizable_type_demotion): Check that types are integral.
1646         (vectorizable_type_promotion): Likewise.
1647         (vectorizable_store): Fix typo.  Eliminate new-line at end of
1648         comments.
1649
1650 2007-01-16  Jan Hubicka  <jh@suse.cz>
1651
1652         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1653         cgraph edges too.
1654
1655 2007-01-15  Eric Christopher  <echristo@apple.com>
1656
1657         * ifcvt.c: Include vec.h, vecprim.h.
1658         (check_cond_move_block): New argument regs.
1659         Reorganize. Add registers used to regs.
1660         (cond_move_process_if_block): Use regs set above as
1661         loop bounds.
1662
1663 2007-01-15  Eric Christopher  <echristo@apple.com>
1664
1665         * config/darwin.h: Update copyright.
1666         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1667         (LINK_COMMAND_SPEC): Add -u.
1668         (LINK_SPEC): Fix umbrella for above.
1669
1670 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
1671
1672         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1673         glibc CVS.
1674
1675 2007-01-15  Tom Tromey  <tromey@redhat.com>
1676
1677         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1678         * doc/install.texi (Testing): Don't mention jacks.
1679         (Configuration): Document --enable-java-maintainer-mode.  Move
1680         --with-java-home to libgcj-specific section.  Document
1681         --with-ecj-jar.
1682         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1683
1684 2007-01-15  Jan Hubicka  <jh@suse.cz>
1685
1686         * tree-ssa-dce.c (DCE_TODOs): New.
1687         (propagate_necessity): Return if something changed.
1688         (eliminate_unnecessary_stmts): Likewise.
1689         (perform_tree_ssa_dce): Return TODO flags when needed.
1690         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1691
1692 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
1693
1694         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1695         (fyl2x_extend<mode>xf3_i387): New insn pattern.
1696         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1697         patterns using X87MODEF12 mode macro.  Extend operand 1
1698         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1699         (log10<mode>2): Ditto.
1700         (log2<mode>2): Ditto.
1701         (log1p<mode>2): Ditto.
1702         (logb<mode>2): Ditto.
1703         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1704         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1705         (*fxtractxf3_i387): Rename from *fxtractxf3.
1706         (fxtract_extend<mode>xf3_i387): New insn pattern.
1707         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1708         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1709         and gen_fyl2xxf3_i387().
1710
1711 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
1712
1713         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1714         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1715         get_loop_exit_edges.  Changed return type to bool.
1716         * cfgloop.h (fix_loop_placement): Declaration removed.
1717
1718 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1719
1720         * param.h (MIN_VECT_LOOP_BOUND): New.
1721         * params.def (MIN_VECT_LOOP_BOUND): New.
1722         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1723         argument - minimum threshold for number of iterations.
1724         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1725         argument to declaration.
1726         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1727         MIN_VECT_LOOP_BOUND.
1728         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1729         slpeel_tree_peel_loop_to_edge with additional argument.
1730         (vect_do_peeling_for_alignment): Likewise.
1731         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1732
1733 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1734
1735         PR target/30413
1736         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1737         operands of size 1.
1738
1739 2007-01-14  Jan Hubicka  <jh@suse.cz>
1740
1741         * tree-dfa.c (remove_referenced_var): New function.
1742         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1743         prune referenced vars list too.
1744         * tree-flow.h (remove_referenced_var): Declare.
1745
1746 2007-01-14  Jan Hubicka  <jh@suse.cz>
1747
1748         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1749         separately.
1750         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1751         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1752         * except.h (verify_eh_throw_table_statements): Kill prototype.
1753         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1754         into call exrepssions.
1755         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1756         eh trhow verifier.
1757         * tree-cfg.c: Include pointer-set.h.
1758         (verify_node_sharing): Work on pointer set.
1759         (verify_eh_throw_stmt_node): New.
1760         (verify_stmts): Use pointers sets, verify throw_stmt.
1761
1762 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1763
1764         * ipa-reference.c (analyze_function): Consider also addresses taken
1765         in phi nodes.
1766
1767 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1768
1769         * c-typeck.c (null_pointer_constant_p): Replace use of
1770         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1771         (build_c_cast): Likewise.
1772
1773 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1774
1775         * tree.h (force_fit_type_double): Remove unused final argument.
1776         * c-common.c (constant_expression_warning): Replace use of
1777         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1778         (convert_and_check): Likewise.
1779         (shorten_compare): Update call to force_fit_type_double.
1780         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1781         * convert.c (convert_to_pointer): Update call to
1782         force_fit_type_double.
1783         * fold-const.c (force_fit_type_double): Remove overflowed_const
1784         argument.
1785         (int_const_binop, fold_convert_const_int_from_int,
1786         fold_convert_const_int_from_real, fold_div_compare,
1787         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1788         fold_abs_const, fold_not_const): Remove the final argument from
1789         calls to force_fit_type_double.
1790
1791 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1792
1793         * configure.ac: Set insn to "nop" for spu-*-* also.
1794         * configure: Regenerate.
1795
1796 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
1797
1798         * builtins.def : Add BUILT_IN_FREE.
1799
1800 2007-01-12  Jan Hubicka  <jh@suse.cz>
1801
1802         PR tree-optimization/30443
1803         * tree-inline.c (tree_function_versioning): Do not optimize when
1804         cloning for inlining.
1805
1806 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
1807
1808         * doc/loop.texi: Document recording of loop exits.
1809         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1810         (update_single_exits_after_duplication,
1811         update_single_exit_for_duplicated_loop,
1812         update_single_exit_for_duplicated_loops): Removed.
1813         (duplicate_loop_to_header_edge): Do not call
1814         update_single_exits_after_duplication and
1815         update_single_exit_for_duplicated_loops.
1816         (loop_version): Do not update single_exit information.
1817         (fix_loop_structure): Use record_loop_exits instead of
1818         mark_single_exit_loops.
1819         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1820         the lists of loop exits.
1821         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1822         split_edge, merge_blocks): Update the lists of loop exits.
1823         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1824         loop_optimizer_init.
1825         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1826         of mark_single_exit_loops.
1827         (loop_optimizer_finalize): Call release_recorded_exits.
1828         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1829         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1830         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1831         update single exit information.
1832         * lambda-code.c (perfect_nestify): Ditto.
1833         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1834         (mark_single_exit_loops): Removed.
1835         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1836         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1837         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1838         functions.
1839         (get_loop_exit_edges, single_exit): Use recorded exit lists.
1840         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1841         (verify_loop_structure): Verify consistency of the exit lists.
1842         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
1843         (set_single_exit): Removed.
1844         * cfgloop.h (struct loop_exit): New function.
1845         (struct loop): single_exit_ field replaced by exits field.
1846         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1847         (struct loops): Added exits hash.
1848         (mark_single_exit_loops, set_single_exit): Declaration removed.
1849         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1850
1851 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1852
1853         * doc/invoke.texi: Avoid use of @headitem.
1854
1855 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1856
1857         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1858
1859 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1860
1861         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1862         list of targets.
1863
1864 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1865             Richard Sandiford  <richard@codesourcery.com>
1866
1867         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1868         * config/m68k/m68k.opt: Resort options.
1869         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1870
1871 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1872             Richard Sandiford  <richard@codesourcery.com>
1873
1874         * doc/invoke.texi: Document the macros that are defined by
1875         m68k's -mtune and -mhard-float options.
1876         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1877         (m68k_cpp_cpu_family): Likewise.
1878         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1879         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
1880         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
1881         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1882         (m68k_cpp_cpu_family): Likewise.
1883
1884 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1885
1886         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1887         macros besides mc68000 as tuning macros.  Use a switch statement
1888         to set them and mcpu32.
1889
1890 2007-01-12  Julian Brown  <julian@codesourcery.com>
1891
1892         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1893         (TARGET_68040_ONLY): Rename to...
1894         (TARGET_68040): ...this.
1895         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1896         * config/m68k/m68k.md: Likewise.
1897
1898 2007-01-12  Julian Brown  <julian@codesourcery.com>
1899             Nathan Sidwell  <nathan@codesourcery.com>
1900             Richard Sandiford  <richard@codesourcery.com>
1901
1902         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1903         the -mcpu= argument associated with the --with-cpu setting.
1904         Define M68K_DEFAULT_TUNE to the default -mtune= option,
1905         if different from the one implied by the -mcpu setting.
1906         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1907         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
1908         * doc/invoke.texi: Mention ColdFire in the introduction to the
1909         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
1910         -mno-div and -mhard-float options.  Make -m68881 a synonym for
1911         -mhard-float.  Document the previously-undocumented -m5206e,
1912         -m528x, -m5307 and -m5407 options.  Tweak the existing option
1913         documentation for consistency.
1914         * doc/install.texi: Mention new --with-cpu arguments.
1915         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1916         default CPU if neither -mcpu nor -march are specified.
1917         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1918         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1919         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
1920         using m68k_tune to decide between families that implement the
1921         same ISA.  Use m68k_tune to set __mcfv4e__.
1922         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1923         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1924         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1925         (FL_ISA_C, FL_ISA_MMU): New macros.
1926         (MASK_COLDFIRE): Delete.
1927         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1928         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1929         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1930         (TARGET_HARD_FLOAT): Do not define here.
1931         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1932         (TUNE_68000): New macro.
1933         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1934         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1935         (TUNE_CFV2): Redefine in terms of m68k_tune.
1936         (uarch_type, target_device, fpu_type): New enums.
1937         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1938         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1939         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1940         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1941         (FL_FOR_isa_c): New macros.
1942         (m68k_isa): New enum.
1943         (m68k_target_selection): New structure.
1944         (all_devices, all_isas, all_microarchs): New tables.
1945         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1946         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1947         (MASK_ALL_CPU_BITS): Delete.
1948         (m68k_find_selection): New function.
1949         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1950         Map the legacy target options to a combination of the new ones.
1951         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1952         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
1953         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1954         settings.
1955         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1956         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1957         properties.
1958         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1959         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1960         * config/m68k/m68k-devices.def: New file.
1961
1962 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1963             Nathan Sidwell  <nathan@codesourcery.com>
1964
1965         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1966         (SUBTARGET_EXTRA_SPECS): New macros.
1967         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1968         use %(asm_cpu_spec) instead.
1969         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1970         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1971         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1972         (EXTRA_SPECS): Rename to...
1973         (SUBTARGET_EXTRA_SPECS): ...this.
1974
1975 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1976             Richard Sandiford  <richard@codesourcery.com>
1977             Julian Brown  <julian@codesourcery.com>
1978
1979         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1980         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1981         to the configuration's default CPU.
1982         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1983         Remove default masks.
1984         (m680[012]0-*-*): Set the default with_cpu to the first part of
1985         the target name.
1986         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1987         (m68k*-*-linux): Extend the --with-cpu handling to...
1988         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
1989         and m68010.  Don't set target_cpu_default2.
1990         * doc/install.texi: Document --with-cpu for m68k.
1991         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1992         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1993         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1994         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1995         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1996         (ASM_SPEC): Remove use of %(asm_cpu_default).
1997         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1998         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1999         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2000         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2001         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2002         (TARGET_DEFAULT): Delete.
2003         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2004         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2005         (CPP_CPU_SPEC): Delete.
2006         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2007         the appropriate string.
2008         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2009         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2010         (ASM_SPEC): Don't use %(asm_default_spec).
2011         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2012         TARGET_DEFAULT and add MASK_68881.
2013         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2014
2015 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2016
2017         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2018         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2019         MASK_68010 alongside MASK_68020.
2020         * doc/invoke.texi: Document -m68010.
2021         * config/m68k/m68k.opt (m68010): New.
2022         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2023         if TUNE_68010.
2024         (TUNE_68010): New macro.
2025         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2026         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2027         (M68K_CPU_m68332): Add MASK_68010.
2028         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2029         fallback definition.
2030         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2031         defines.
2032         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2033         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2034         to all entries that use MASK_68020.
2035         (output_move_simode_const, output_move_himode, output_move_qimode)
2036         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2037         instead of TARGET_68020 to select clr behavior.  Remove comment
2038         about there being no TARGET_68010.
2039         * config/m68k/m68k.md: Likewise throughout.
2040
2041 2007-01-12  Julian Brown  <julian@codesourcery.com>
2042
2043         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2044         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2045         * config/m68k/m68k.md: Likewise.
2046
2047 2007-01-12  Julian Brown  <julian@codesourcery.com>
2048
2049         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2050         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2051
2052 2007-01-12  Julian Brown  <julian@codesourcery.com>
2053
2054         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2055         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2056         * config/m68k/m68k.md: Likewise.
2057
2058 2007-01-12  Julian Brown  <julian@codesourcery.com>
2059             Richard Sandiford  <richard@codesourcery.com>
2060
2061         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2062         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2063         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2064         TARGET_CPU32.
2065         (TARGET_CPU32): Rename to...
2066         (TUNE_CPU32): ...this.
2067         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2068         (TUNE_CFV2): New macros.
2069         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2070         remove conditions that are implied by TARGET_68020.
2071         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2072         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2073         (m68k_output_function_epilogue): Likewise.
2074         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2075         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2076         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2077         non-68000 timings.  Refactor multiplication and division costs.
2078         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2079         TUNE_CPU32 instead of TARGET_CPU32.
2080         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2081         and TUNE_68060 instead of TARGET_68060.
2082         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2083         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2084         TARGET_CPU32.
2085         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2086         && !TARGET_COLDFIRE" to choose between moveq and clr.
2087         Likewise in the unnamed movsf pattern.
2088         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2089         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2090         ashiftrt pattern.
2091
2092 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2093
2094         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2095         of tabbing before backslashes.
2096
2097 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2098
2099         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2100         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2101         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2102
2103 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2104
2105         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2106         find_interesting_uses_cond.
2107         (find_interesting_uses_cond): Use extract_cond_operands.
2108         (rewrite_use_compare): Use extract_cond_operands and
2109         force_gimple_operand_bsi.  Do not call update_stmt.
2110         (determine_use_iv_cost_condition): Use extract_cond_operands.
2111         Return cheaper of using original bound and changing the exit bound.
2112
2113 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2114
2115         PR tree-optimization/29516
2116         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2117         most_expensive_mult_to_index, addr_to_parts,
2118         create_mem_ref, maybe_fold_tmr): Make the type of
2119         fields of TARGET_MEM_REF sizetype.
2120         (move_fixed_address_to_symbol, move_pointer_to_base):
2121         New functions.
2122         * tree.def (TARGET_MEM_REF): Add comment on types of
2123         the operands.
2124
2125 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
2126
2127         * c-common.c (vector_types_convertible_p): Treat opaque types as
2128         always convertible if they have the same size, but not otherwise.
2129
2130 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
2131
2132         * ifcvt.c (struct noce_if_info): Add comments to the fields.
2133         Remove the b_unconditional field.
2134         (noce_try_sign_mask): Do not look at b_unconditional.
2135         (noce_process_if_block): Do not use merge_if_blocks.  Update
2136         the CFG here.  Do not set b_unconditional.
2137         (cond_move_process_if_block): Likewise.
2138         (find_cond_trap): Likewise.
2139         (check_cond_move_block): Require simple jump insns at the end
2140         of the basic block.
2141
2142 2007-01-11  Jan Hubicka  <jh@suse.cz>
2143
2144         PR tree-optimization/1046
2145         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2146         when alias info is not ready.
2147         (pass_tail_recursion): Do not require aliasing.
2148         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2149         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2150         * tree-ssa-copy.c (pass_copy_prop): Likewise.
2151         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2152         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2153         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2154         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2155         before inlining.
2156         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2157         When aliasing is not build, mark statement as volatile.
2158
2159 2007-01-11  Tom Tromey  <tromey@redhat.com>
2160
2161         PR preprocessor/15185, PR preprocessor/20989:
2162         * doc/cppopts.texi <-MT>: Update description of algorithm for
2163         computing default target.
2164         <-M, -MD>: Reword "basename" text.
2165
2166 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2167
2168         * builtins.c (expand_builtin_pow, expand_builtin_powi,
2169         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2170         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2171         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2172         fold_builtin_bswap, real_constp, fold_builtin_pow,
2173         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2174         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2175         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2176         * convert.c (convert_to_pointer): Likewise.
2177         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2178         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2179         const_binop, fold_convert_const_int_from_int,
2180         fold_convert_const_int_from_real,
2181         fold_convert_const_real_from_real, sign_bit_p,
2182         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2183         fold_sign_changed_comparison, fold_unary, fold_comparison,
2184         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2185         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2186         * print-tree.c (print_node_brief, print_node): Likewise.
2187         * stor-layout.c (place_field, layout_type): Likewise.
2188         * tree-chrec.c (keep_cast): Likewise.
2189         * tree.c (build_vector, build_real, build_real_from_int_cst,
2190         build_complex): Likewise.
2191
2192 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2193
2194         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
2195         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2196
2197 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
2198
2199         * configure.ac (strict1_warn): Rename to strict_warn.
2200         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2201         cc_set_by_configure, quoted_cc_set_by_configure,
2202         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2203         all_boot_languages, all_stagestuff): Remove.
2204         (target_list): Remove bootstrap targets.
2205         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2206         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2207         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2208         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2209         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2210         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2211         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2212         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2213         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2214         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2215         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2216         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2217         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2218         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2219         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2220         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2221         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2222         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2223         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2224         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2225         (mostlyclean): Adjust.
2226         (clean, distclean): Don't mention bootstrap stuff.
2227         * configure: Regenerate.
2228         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2229         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2230         treelang/config-lang.in (stagestuff): Remove.
2231         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2232         stageprofile, stagefeedback, stagestuff): Remove mention.
2233
2234 2007-01-11  Nick Clifton  <nickc@redhat.com>
2235
2236         * config/mcore/predicates.md (mcore_general_movesrc_operand):
2237         Accept CONSTs.
2238         (mcore_general_movdst_operand): Do not accept CONST_INTs.
2239         (mcore_arith_K_S_operand): Run the test for the S constraint not
2240         the test for the M constraint.
2241         (mcore_addsub_operand): Do not accept integer values that are
2242         larger than 32 bits.
2243         * config/mcore/mcore.md: Remove unused constraints from split.
2244         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2245         (addsi3): Likewise.
2246         (allocate_stack): Likewise.
2247         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2248         operands to 32 bits.
2249         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2250         INTVAL.
2251         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2252         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2253         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2254         mcore_output_bclri, mcore_output_andn, output_inline_const,
2255         mcore_output_move, mcore_output_movedouble): Likewise.
2256         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2257         (output_inline_const): Likewise.
2258         (output_inline_const): Fix format strings used in sprintf
2259         statements.
2260         * config/mcore/mcore-protos.h: Update prototypes for changed
2261         functions in mcore.c.
2262         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2263         HOST_WIDE_INT and not int.
2264         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2265         CONST_OK_FOR_N): Likewise.
2266         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2267         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2268         an INTVAL.
2269
2270 2007-01-10  Jan Hubicka  <jh@suse.cz>
2271
2272         * tree-vrp.c (remove_range_assertions): Release defs.
2273         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2274         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2275
2276 2007-01-10  Paul Brook  <paul@codesourcery.com>
2277
2278         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2279         * config/arm/arm.md (mulsi3subsi): New insn.
2280
2281 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
2282
2283         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2284         (tree_transform_and_unroll_loop): New.
2285         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2286         Declare.
2287
2288 2007-01-10  Robert Kennedy <jimbob@google.com>
2289
2290         * fold-const.c (fold_comparison): Fold comparisons like (x *
2291         1000 < 0) to (x < 0).
2292
2293 2007-01-10  Ian Lance Taylor  <iant@google.com>
2294
2295         * tree-pretty-print.c (dump_generic_node): Print parentheses when
2296         operands have the same priority.
2297
2298 2007-01-10  Tom Tromey  <tromey@redhat.com>
2299
2300         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2301         (fold_binary): Likewise.
2302         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2303         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2304         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2305         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2306         * langhooks.h (struct lang_hooks): Removed field
2307         'can_use_bit_fields_p'.
2308
2309 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
2310
2311         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2312
2313 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
2314
2315         * function.c (get_last_funcdef_no): New function.
2316         * function.h (get_last_funcdef_no): Declare.
2317         * tree-inline.c (initialize_cfun): Add initialization.
2318         (tree_function_versioning): Cleanup.
2319
2320 2007-01-10  Jan Hubicka  <jh@suse.cz>
2321
2322         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2323         PHIs.
2324
2325 2007-01-10  Sa Liu  <saliu@de.ibm.com>
2326             Ben Elliston  <bje@au.ibm.com>
2327
2328         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2329         for SAVE_NONLOCAL and Pmode for any other save level.
2330         * spu-protos.h (spu_restore_stack_block): Declare.
2331         * spu.md (save_stack_block): Remove.
2332         (restore_stack_block): Call spu_restore_stack_block.
2333         * spu.c (spu_restore_stack_block): New function.
2334         (spu_expand_epilogue): Remove old comment.
2335
2336 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
2337
2338         PR tree-optimization/30322
2339         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2340         (cand_value_at): Return the value as aff_tree.
2341         (may_eliminate_iv): Convert the bound from aff_tree to tree.
2342         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2343         aff_combination_mult): New functions.
2344         (aff_combination_add): Use aff_combination_add_cst.
2345         (aff_combination_convert): Allow conversions to a wider type.
2346         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2347         * tree-affine.h (aff_combination_mult): Declare.
2348
2349 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
2350
2351         * doc/tm.texi: Update documentation to reflect reality of exec
2352         and start file search behaviours. Update copyright year.
2353         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2354         header file directories.
2355
2356 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
2357
2358         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2359         (*cosxf2): Rename to cosxf2_i387.
2360         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
2361         using X87MODEF12 mode macro. Rename patterns to
2362         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
2363         for SSE math.
2364         (*cosdf2, *cossf2): Ditto.
2365         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
2366         to match extended input operands.
2367         (sincos<mode>3): New expander.
2368         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2369         insn patterns and corresponding splitters.
2370
2371 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
2372
2373         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2374         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2375
2376 2007-01-09  Nicolas Pitre  <nico@cam.org>
2377
2378         PR target/30173
2379         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2380
2381 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2382
2383         * target.h (struct gcc_target): New field library_rtti_comdat.
2384         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2385         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2386         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2387         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2388
2389 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2390
2391         * doc/invoke.texi (Optimize Options): Correct description of -O0.
2392
2393 2007-01-08  Richard Guenther  <rguenther@suse.de>
2394
2395         * tree.h (force_fit_type_double): Export.
2396         (force_fit_type): Remove.
2397         * fold-const.c (force_fit_type_double): New function.
2398         (force_fit_type): Remove.
2399         (int_const_binop): Use it.
2400         (fold_convert_const_int_from_int): Likewise.
2401         (fold_convert_const_int_from_real): Likewise.
2402         (fold_div_compare): Likewise.
2403         (fold_sign_changed_comparison): Likewise.
2404         (fold_unary): Likewise.
2405         (fold_negate_const): Likewise.
2406         (fold_abs_const): Likewise.
2407         (fold_not_const): Likewise.
2408         * c-common.c (shorten_compare): Use force_fit_type_double.
2409         * convert.c (convert_to_pointer): Likewise.
2410
2411 2007-01-08  Richard Guenther  <rguenther@suse.de>
2412
2413         * tree.h (build_int_cst_wide_type): Export.
2414         * tree.c (build_int_cst_wide_type): New function.
2415         (build_int_cst_wide): Fix comment.
2416         * builtins.c (fold_builtin_object_size): Use build_int_cst
2417         to build -1 or 0 of the correct type.  Use fit_double_type
2418         to check for overflow.
2419         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2420         to build the mask.
2421         (decode_field_reference): Likewise.
2422         (all_ones_mask_p): Likewise.
2423         (native_interpret_int): Use build_int_cst_wide_type.
2424         (fold_binary): Use build_int_cst_type to build an all-ones
2425         value.
2426         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2427
2428 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
2429
2430         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2431
2432 2007-01-09  Ben Elliston  <bje@au.ibm.com>
2433
2434         * genautomata.c (STATS_OPTION): New option.
2435         (stats_flag): New flag.
2436         (gen_automata_option): Handle it.
2437         (initiate_automaton_gen): Ditto.
2438         (write_automata): Output statistics only if stats_flag is
2439         set. Likewise, output time statistics only if time_flag is set.
2440         * doc/md.texi (Processor pipeline description): Document new flag.
2441
2442 2007-01-08  Richard Guenther  <rguenther@suse.de>
2443
2444         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2445         * tree.c (build_int_cst_type): Likewise.
2446         (size_in_bytes): Don't call force_fit_type on the result.
2447         (int_fits_type_p): Use fit_double_type.
2448         * fold-const.c (fit_double_type): New function.
2449         (force_fit_type): Use it.
2450         * tree.h (fit_double_type): Export.
2451
2452 2007-01-08  Jan Hubicka  <jh@suse.cz>
2453
2454         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2455         * ipa.c (function_and_variable_visibility): Fix return type.
2456
2457 2007-01-08  Richard Guenther  <rguenther@suse.de>
2458
2459         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2460         of offset to build the index.
2461         * tree-pretty-print.c (dump_generic_node): Don't build negated
2462         const just for printing.
2463         * c-pretty-print.c (pp_c_integer_constant): Likewise.
2464         * builtins.c (fold_builtin_int_roundingfn): Check if result
2465         fits the type by using force_fit_type and comparing the result.
2466         * predict.c (predict_loops): Use compare_tree_int for comparison.
2467         * tree.c (build_int_cst): Fall back to integer_type_node for
2468         NULL_TREE type.
2469         (build_int_cst_wide): Assert type is non-null.
2470
2471 2007-01-08  Roberto Costa  <roberto.costa@st.com>
2472
2473         * tree-vrp.c (extract_range_from_cond_expr): New.
2474         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2475         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2476         as expressions.
2477         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2478         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2479         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2480         used as expressions.
2481         * tree-object-size.c (cond_expr_object_size): New.
2482         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2483
2484 2007-01-08  Jan Hubicka  <jh@suse.cz>
2485
2486         * tree-ssa-forwprop.c (forward_propagate_into_cond,
2487         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2488         statement.
2489
2490 2007-01-08  Richard Guenther  <rguenther@suse.de>
2491
2492         PR tree-optimization/23603
2493         * tree-vrp.c (set_value_range_to_truthvalue): New function.
2494         (extract_range_from_binary): Fall back to truthvalue instead of
2495         varying for TRUTH_*_EXPR.
2496         (extract_range_from_comparison): Fall back to truthvalue instead of
2497         varying.
2498         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2499         if all visited PHI values were constant.
2500
2501 2007-01-08  Jan Hubicka  <jh@suse.cz>
2502
2503         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2504         (cgraph_analyze_function): break out from ...
2505         (cgraph_finalize_compilation_unit): ... here.
2506         (cgraph_expand_function): Remove forgoten commented out line.
2507         (cgraph_optimize): Analyze functions.
2508
2509 2007-01-08  Jan Hubicka  <jh@suse.cz>
2510
2511         * tree-pas.h (TODO_remove_function): New flag.
2512         (TODO_update*): Renumber.
2513         (pass_ipa_increase_alignment,
2514         pass_ipa_function_and_variable_visibility): New passes.
2515         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2516         (cgraph_function_and_variable_visibility): Move to ipa.c
2517         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2518         cgraph_increase_alignment.
2519         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2520         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2521         calling tree-inline.
2522         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2523         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2524         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2525         (gate_increase_alignment): New function.
2526         (pass_ipa_increase_alignment): New pass.
2527         * ipa.c: Inline tree-pass.h and timevar.h
2528         (function_and_variable_visibility): Move here from cgraphunit.c
2529         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2530         * passes.c (init_optimization_passes): Add the two new passes.
2531         (execute_todo): Handle cgraph_remove_functions.
2532
2533 2007-01-08  Nick Clifton  <nickc@redhat.com>
2534
2535         * config/frv/predicates.md (reg_or_0_operand): Accept
2536         CONST_DOUBLEs.
2537
2538 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
2539
2540         * config/bfin/rtems.h, config/bfin/t-rtems: New.
2541         * config.gcc: Add bfin*-rtems*.
2542
2543 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2544
2545         * c.opt: Add -flax-vector-conversions.
2546         * c-typeck.c (convert_for_assignment): Pass flag to
2547         vector_types_convertible_p to allow emission of note.
2548         (digest_init): Likewise.
2549         * c-opts.c: Handle -flax-vector-conversions.
2550         * c-common.c (flag_lax_vector_conversions): New.
2551         (vector_types_convertible_p): Unless -flax-vector conversions
2552         has been passed, disallow conversions between vectors with
2553         differing numbers of subparts and/or element types.  If such
2554         a conversion is disallowed, possibly emit a note on the first
2555         occasion only to inform the user of -flax-vector-conversions.
2556         The new last argument specifies this.
2557         * c-common.h (flag_lax_vector_conversions): New.
2558         (vector_types_convertible_p): Add extra argument.
2559         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2560         char_type_node for V*QI type vectors.
2561         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2562         Update to satisfy new typechecking rules.
2563         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2564         C and C++ variants.
2565         * doc/invoke.texi (C Dialect Options): Document
2566         -flax-vector-conversions.
2567
2568 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2569
2570         PR tree-optimization/29877
2571         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2572         a register variable on the RHS to not be replaceable.
2573
2574 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
2575         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2576         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2577         const_simm15, const_pow2, const_npow2): Added.
2578         * config/score/misc.md (insv, extv, extzv, movmemsi,
2579         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2580         * config/score/score.c (score_address_cost, score_select_cc_mode):
2581         Added.
2582         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2583         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2584         Update score_print_operand.
2585         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2586         Adjust register allocate order and update some macro define.
2587         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2588         mdx_block_move_straight, mdx_block_move_loop_head,
2589         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2590         mdx_block_move): Added.
2591         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2592         mdp_limm): Updated and fix some bug and typo.
2593         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2594         (movsf, movdf, doloop_end): Added.
2595
2596 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
2597
2598         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2599         config/arm/thumb2.md: Fix comment typos.
2600         * doc/extend.texi: Fix a typo.
2601
2602 2007-01-07  Eric Christopher  <echristo@apple.com>
2603
2604         * configure.ac: Check for __stack_chk_fail for darwin.
2605         * configure: Regenerate.
2606
2607 2007-01-07  Richard Guenther  <rguenther@suse.de>
2608
2609         * tree-vrp.c (extract_range_from_assert): CSE calls to
2610         compare_values where possible.
2611         (extract_range_from_unary_expr): Likewise.
2612
2613 2007-01-07  Anatoly Sokolov <aesok@post.ru>
2614
2615         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2616         * config/avr/avr.c (call_insn_operand): Delete function.
2617         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2618         instead of register number. Use predicates.
2619         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2620         call_insn_operand): Add.
2621
2622 2007-01-06  Jan Hubicka  <jh@suse.cz>
2623
2624         * tree-pass.h (pass_build_cgraph_edges): Declare.
2625         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2626         (visited_nodes): Remove.
2627         (cgraph_create_edges): Move to cgraphbuild.c; rename to
2628         build_cgrpah_edges; make visited_nodes local.
2629         (cgraph_process_new_functions): DO not call initialize_inline_failed.
2630         (record_references_in_initializer): Move to cgraphbuild.c
2631         (initialize_inline_failed, rebuild_cgraph_edges,
2632         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2633         (verify_cgraph_node): Make visited_nodes local.
2634         (cgraph_analyze_function): Do not call cgraph_create_edges and
2635         initialize_inline_failed.
2636         (cgraph_expand_function): Do not call cgraph_lower_function;
2637         assert that function is already lowered.
2638         * Makefile.in (cgraphbuild.o): New.
2639         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2640         at the end of lowering passes.
2641
2642 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
2643
2644         * ifcvt.c (cond_move_convert_if_block): New function, code
2645         factored out from...
2646         (cond_move_process_if_block): ...here.  Call the new function
2647         on the THEN and ELSE blocks.
2648         (merge_if_block): Do not copy global_live_at_end, merge_blocks
2649         already takes care of this.
2650
2651 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2652
2653         PR c/19978
2654         * tree.h (TREE_OVERFLOW_P): New.
2655         * c-typeck.c (parser_build_unary_op): Warn only if result
2656         overflowed and operands did not.
2657         (parser_build_binary_op): Likewise.
2658         (convert_for_assignment): Remove redundant overflow_warning.
2659         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2660
2661 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2662
2663         * c-typeck.c (store_init_value): Split over two lines to follow
2664         the GNU coding style.
2665
2666 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
2667
2668         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2669         __GXX_EXPERIMENTAL_CXX0X__.
2670         * doc/cpp.texi: Same.
2671
2672 2007-01-05  Richard Guenther  <rguenther@suse.de>
2673
2674         PR middle-end/27826
2675         * tree.c (get_narrower): Do not construct COMPONENT_REFs
2676         with mismatched types.  Instead explicitly build a
2677         conversion NOP_EXPR.
2678
2679 2007-01-05  Ian Lance Taylor  <iant@google.com>
2680
2681         * c-common.c (decl_with_nonnull_addr_p): New function.
2682         (c_common_truthvalue_conversion): Call it.
2683         * c-typeck.c (build_binary_op): Likewise.
2684         * c-common.h (decl_with_nonnull_addr_p): Declare.
2685
2686 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
2687
2688         PR c/30360
2689         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2690         denom against 0.0.
2691
2692 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
2693
2694         * doc/install.texi (Final install): Document the fact that
2695         the GNAT runtime should not be stripped.
2696
2697 2007-01-04  Jan Hubicka  <jh@suse.cz>
2698
2699         * tree-inline.c (fold_marked_statements): Update operand caches
2700         and EH after folding
2701
2702 2007-01-04  Ian Lance Taylor  <iant@google.com>
2703
2704         * c-common.c (check_function_nonnull): Whitespace fix.
2705
2706 2007-01-04  Jan Hubicka  <jh@suse.cz>
2707
2708         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2709         comitted older version of patch.
2710         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2711         function, celanup cfg and collect garbage.
2712
2713 2007-01-04  Mike Stump  <mrs@apple.com>
2714
2715         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2716         (clean): Likewise.
2717
2718 2007-01-04  Eric Christopher  <echristo@apple.com>
2719
2720         * libgcc2.c (__bswapsi2): Use SItype.
2721         (__bswapdi2): Use DItype.
2722         * libgcc2.h: Update for above.
2723
2724 2007-01-04  Paul Brook  <paul@codesourcery.com>
2725
2726         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2727         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2728         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2729         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2730         umulsi3_highpart): Make conditional on !arm_arch6.
2731         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2732         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2733         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2734         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2735         umulsi3_highpart_v6): New insns.
2736
2737 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2738
2739         * fold-const.c (fold_convert): When casting an expression to void,
2740         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2741         have a type.  Instead of attempting to build a NOP_EXPR, return
2742         these "special" trees directly.
2743
2744 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2745
2746         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2747         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2748         rs6000_cost->fp not 0 as adjustment for outer NEG.
2749
2750 2007-01-04  Jan Hubicka  <jh@suse.cz>
2751
2752         * cgraph.c (cgraph_release_function_body): New function.
2753         (cgraph_remove_node): Use it.
2754         * cgraph.h (cgraph_release_function_body): Declare.
2755         * cgraphunit.c (cgraph_expand_function): Use it.
2756         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2757         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2758         is initialized and while compilation of other function is running.
2759         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2760         statement CFG annotations from here to ...
2761         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2762
2763 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2764
2765         * cfgloop.h (enum li_flags): Make the constants powers of two.
2766
2767 2007-01-04  Jan Hubicka  <jh@suse.cz>
2768
2769         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2770         set.
2771         (fold_marked_statements): New function.
2772         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2773         * tree-inline.h (copy_body_data): Add statements_to_fold.
2774
2775 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2776
2777         * config.gcc: Mention libgcc/config.host.
2778         * Makefile.in: Update comments mentioning libgcc.
2779         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2780         (all.cross, start.encap, rest.encap, rest.cross): Update
2781         dependencies for libgcc move.
2782         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2783         (clean-target, clean-target-libgcc): Delete.
2784         (srcdirify, GCC_EXTRA_PARTS): New macros.
2785         (libgcc-support, libgcc.mvars): New rules.
2786         (distclean): Remove mention of mklibgcc.
2787         (install): Don't reference INSTALL_LIBGCC.
2788         (install-common): Don't reference EXTRA_PARTS.
2789         (install-libgcc, install-multilib): Delete rules.
2790         * mklibgcc.in: Delete file.
2791         * doc/configfiles.texi: Don't mention mklibgcc.
2792
2793         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2794         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2795         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2796         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2797         (mklibgcc, ldblspecs): Likewise.
2798
2799         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2800         (posixpre.def): Use $(T).
2801         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2802         (SHLIB_SRC, SHLIB_INSTALL): Delete.
2803         (SHLIB_LINK): Make dummy.
2804         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2805
2806         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2807
2808         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2809         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2810         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2811         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2812
2813         * configure.ac (all_outputs): Remove mklibgcc.
2814         * configure: Regenerated.
2815
2816 2007-01-03  Josh Conner  <jconner@apple.com>
2817
2818         PR middle-end/29683
2819         * calls.c (compute_argument_addresses): Set stack and stack_slot
2820         for partial args, too.
2821         (store_one_arg): Use locate.size.constant for the size when
2822         generating a save_area.
2823
2824 2007-01-03  Robert Kennedy <jimbob@google.com>
2825
2826         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2827         whose definitions are deleted due to basic block merging.
2828
2829 2007-01-03  Paul Brook  <paul@codesourcery.com>
2830
2831         PR target/16634
2832         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2833         functions.
2834         (use_return_insn): Return 0 for Thumb interrupt functions.
2835         (print_multi_reg): Add rfe argument for IRQ returns.
2836         (arm_output_epilogue): Pop interrupt return address directly into PC.
2837         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2838
2839 2007-01-03  Paul Brook  <paul@codesourcery.com>
2840
2841         Merge from sourcerygxx-4_1.
2842         * config/arm/thumb2.md: New file.
2843         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2844         Thumb-2.
2845         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2846         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2847         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2848         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2849         tables.
2850         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2851         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2852         comptibility.
2853         * config/arm/ieee754-sf.S: Ditto.
2854         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2855         (thumb1_base_register_rtx_p): ... to this.
2856         (thumb_index_register_rtx_p): Rename...
2857         (thumb1_index_register_rtx_p): ... to this.
2858         (thumb_output_function_prologue): Rename...
2859         (thumb1_output_function_prologue): ... to this.
2860         (thumb_legitimate_address_p): Rename...
2861         (thumb1_legitimate_address_p): ... to this.
2862         (thumb_rtx_costs): Rename...
2863         (thumb1_rtx_costs): ... to this.
2864         (thumb_compute_save_reg_mask): Rename...
2865         (thumb1_compute_save_reg_mask): ... to this.
2866         (thumb_final_prescan_insn): Rename...
2867         (thumb1_final_prescan_insn): ... to this.
2868         (thumb_expand_epilogue): Rename...
2869         (thumb1_expand_epilogue): ... to this.
2870         (arm_unwind_emit_stm): Rename...
2871         (arm_unwind_emit_sequence): ... to this.
2872         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2873         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2874         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2875         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2876         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2877         thumb2_output_casesi): New functions.
2878         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2879         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2880         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2881         THUMB2_WORK_REGS): Define.
2882         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2883         arm_condexec_mask, arm_condexec_masklen)): New variables.
2884         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2885         (arm_override_options): Check new CPU capabilities.
2886         Set new architecture flag variables.
2887         (arm_isr_value): Handle v7m interrupt functions.
2888         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
2889         Thumb-2.
2890         (const_ok_for_arm): Handle Thumb-2 constants.
2891         (arm_gen_constant): Ditto.  Use movw when available.
2892         (arm_function_ok_for_sibcall): Return false for v7m interrupt
2893         functions.
2894         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2895         (thumb_find_work_register, arm_load_pic_register,
2896         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2897         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2898         print_multi_reg, output_mov_long_double_fpa_from_arm,
2899         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2900         output_mov_double_fpa_from_arm, output_move_double,
2901         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2902         output_return_instruction, arm_output_function_prologue,
2903         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2904         arm_output_mi_thunk, thumb_set_return_address): Ditto.
2905         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
2906         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2907         (arithmetic_instr, shift_op): Use arm_shift_nmem.
2908         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
2909         '!' and 'L'.
2910         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2911         (thumb_expand_prologue): Use thumb_set_frame_pointer.
2912         (arm_file_start): Output directive for unified syntax.
2913         (arm_unwind_emit_set): Handle stack alignment instruction.
2914         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2915         Add v6t2, v7, v7a, v7r and v7m.
2916         (RETLDM): Add Thumb-2 code.
2917         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2918         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2919         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2920         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2921         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2922         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2923         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2924         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2925         ADJUST_INSN_LENGTH): Define.
2926         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2927         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2928         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2929         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2930         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2931         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2932         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2933         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2934         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2935         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2936         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2937         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2938         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2939         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2940         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2941         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2942         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2943         cortex-m3.
2944         * config/arm/arm-tune.md: Regenerate.
2945         * config/arm/arm-protos.h: Update prototypes.
2946         * config/arm/vfp.md: Enable patterns for Thumb-2.
2947         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
2948         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2949         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2950         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2951         * config/arm/libunwind.S: Add Thumb-2 code.
2952         * config/arm/constraints.md: Update include Thumb-2.
2953         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2954         * config/arm/ieee754-df.S: Ditto.
2955         * config/arm/bpabi.S: Ditto.
2956         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2957         * config/arm/predicates.md (low_register_operand,
2958         low_reg_or_int_operand, thumb_16bit_operator): New.
2959         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2960         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2961         * config/arm/t-arm-elf: Add armv7 multilib.
2962         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2963         Include thumb2.md.
2964         (UNSPEC_STACK_ALIGN, ce_count): New.
2965         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2966         arm_zero_extendsidi2, arm_zero_extendqidi2): New
2967         insns/expanders.
2968         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2969         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2970         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2971         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2972         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2973         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2974         insns.
2975         * doc/extend.texi: Document ARMv7-M interrupt functions.
2976         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2977
2978 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
2979
2980         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2981         (struct _Unwind_Context): Rename args_size to flags, remove
2982         signal_frame field, add a new args_size field and version field.
2983         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2984         _Unwind_IsExtendedContext): New inline functions.
2985         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2986         Assume by_value array is only present if _Unwind_IsExtendedContext.
2987         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2988         _Unwind_IsSignalFrame.
2989         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2990         (uw_update_context_1): Use _Unwind_SetSignalFrame.
2991         (uw_init_context_1): Initialize context->flags to
2992         EXTENDED_CONTEXT_BIT.
2993         * config/rs6000/linux-unwind.h (frob_update_context): Use
2994         _Unwind_SetSignalFrame.
2995
2996 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
2997
2998         PR middle-end/30353
2999         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3000         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3001         the value.
3002
3003 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3004
3005         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3006         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3007         RET, AT_SP): Remove.
3008         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3009         instead of RET.
3010
3011         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3012         * cfgbuild.c (find_basic_blocks): Likewise.
3013         * cfgrtl.c (rtl_create_basic_block): Likewise.
3014         * function.c (temp_slots_at_level): Likewise.
3015         * reg-stack.c (stack_regs_mentioned): Likewise.
3016         * regclass.c (allocate_reg_info): Likewise.
3017         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3018         set_bb_for_stmt, move_block_to_fn): Likewise.
3019         * tree-complex.c (tree_lower_complex): Likewise.
3020         * vec.h (VEC_safe_grow_cleared): New.
3021
3022         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3023
3024 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
3025
3026         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3027         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3028         to loop_version.
3029         * cfgloopmanip.c (scale_loop_frequencies): Export.
3030         (loopify): Scale the frequencies by prescribed coefficients.
3031         (set_zero_probability): New function.
3032         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3033         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3034         and frequencies according to arguments.
3035         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3036         correctly.
3037         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3038         * modulo-sched.c (sms_schedule): Set probabilities for entering
3039         versioned loop correctly.
3040         * tree-vect-transform.c (vect_transform_loop): Ditto.
3041         * cfgloop.h (loopify, loop_version): Declaration changed.
3042         (scale_loop_frequencies): Declared.
3043
3044 2007-01-02  Jan Hubicka  <jh@suse.cz>
3045
3046         * cgraph.c: Include tree-flow.h
3047         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3048         early_local_passes.
3049         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3050         * tree-pass.h (pass_all_early_optimizations): Declare.
3051         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3052         early_local_passes.
3053         (cgraph_analyze_function): Do early_local_passes.
3054         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3055         Do not add referenced vars.
3056         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3057         SSA form.
3058         (gate_all_early_local_passes): New gate.
3059         (pass_early_local_passes): Use new gate.
3060         (execute_early_local_optimizations): New functions.
3061         (gate_all_early_optimizations): New gate.
3062         (pass_all_early_optimizations): New pass.
3063         (execute_free_datastructures): Free SSA only when initialized.
3064         (gate_init_datastructures): Init only when optimizing.
3065         (tree_lowering_passes): Do early local passes when called late.
3066         * tree-profile.c (do_tree_profiling): Don't profile functions added
3067         late.
3068         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3069         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3070         not active.
3071         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3072         during early_local_passes.
3073         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3074
3075
3076 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3077
3078         * Makefile.in: Update copyright year.
3079
3080 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3081
3082         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3083         in $(lang_checks) and check-consistency targets.
3084
3085 2007-01-02  Jan Hubicka  <jh@suse.cz>
3086
3087         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3088         Do not add referenced vars.
3089         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3090         are not active.
3091         * passes.c (init_optimization_passes): Put mudflap_2 after
3092         free_datastructures.
3093
3094 2007-01-02  Jan Hubicka  <jh@suse.cz>
3095
3096         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3097         Set NOTHROW flag on call statements proved to be nothrow.
3098         Update statement of local calls so new pure/const functions are
3099         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3100         goto receivers.
3101         (tree_rest_of_compilation): Register hooks and initialize bitmap
3102         early. Do not set after_inlining flag.
3103
3104 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3105
3106         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3107         HOST_BITS_PER_LONGLONG
3108
3109 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3110
3111         PR c/19977
3112         * c-typeck.c (store_init_value): Don't emit pedantic overflow
3113         warning for non-static initializers.
3114
3115 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
3116
3117         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3118         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3119         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3120         Always use set_unique_reg_note to add REG_EQUAL notes.
3121
3122 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3123
3124         Revert:
3125         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3126
3127         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3128         * cfgbuild.c (find_basic_blocks): Likewise.
3129         * cfgrtl.c (rtl_create_basic_block): Likewise.
3130         * function.c (temp_slots_at_level): Likewise.
3131         * reg-stack.c (stack_regs_mentioned): Likewise.
3132         * regclass.c (allocate_reg_info): Likewise.
3133         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3134         set_bb_for_stmt, move_block_to_fn): Likewise.
3135         * tree-complex.c (tree_lower_complex): Likewise.
3136         * vec.h (VEC_safe_grow_cleared): New.
3137
3138 2007-01-02  Ian Lance Taylor  <iant@google.com>
3139
3140         * c-common.c (c_common_truthvalue_conversion): When warning about
3141         using an assignment as a truth value, set TREE_NO_WARNING.
3142
3143 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3144
3145         PR middle-end/7651
3146         * c.opt (Wold-style-declaration): New.
3147         * doc/invoke.texi (C-only Warning Options): New.
3148         (Wold-style-declaration): Document it.
3149         (Wextra): Enabled by -Wextra.
3150         * c-opts.c (c_common_post_options): Enabled by -Wextra.
3151         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3152         -Wold-style-declaration.
3153
3154 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3155
3156         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3157         * cfgbuild.c (find_basic_blocks): Likewise.
3158         * cfgrtl.c (rtl_create_basic_block): Likewise.
3159         * function.c (temp_slots_at_level): Likewise.
3160         * reg-stack.c (stack_regs_mentioned): Likewise.
3161         * regclass.c (allocate_reg_info): Likewise.
3162         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3163         set_bb_for_stmt, move_block_to_fn): Likewise.
3164         * tree-complex.c (tree_lower_complex): Likewise.
3165         * vec.h (VEC_safe_grow_cleared): New.
3166
3167 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
3168
3169         * c-common.c (c_common_nodes_and_builtins): Since variants of
3170         void_type_node get built before it is given a name, we need to
3171         give those variants the name, too.
3172         (complete_array_type): We need to work with the canonical main
3173         type of the array, from which we will build the qualified version.
3174         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3175         * print-tree.c (print_node): Display canonical type information
3176         for each type.
3177         * stor-layout.c (layout_type): When we don't know the
3178         alignment of a type for which we're building an array, we end up
3179         guessing wrong, so make the type require structural equality.
3180         * tree.c (make_node_stat): When we build a new type, it is its
3181         own canonical type.
3182         (build_type_attribute_qual_variant): When building an attribute
3183         variant, its canonical type is the non-attribute variant. However,
3184         if the attributes are target-dependent and they differ, we need to
3185         use structural equality checks for this type.
3186         (build_qualified_type): A qualified type is not equivalent to its
3187         unqualified variant; set the canonical type appropriately.
3188         (build_distinct_type_copy): When building a distinct type from
3189         another type, the new type is its own canonical type.
3190         (build_variant_type_copy): When building a new type variant, we
3191         assume that it is equivalent to the original type.
3192         (build_pointer_type_for_mode): When building a pointer type, also
3193         build a canonical type pointer.
3194         (build_reference_type_for_mode): When building a reference type,
3195         also build a canonical type reference.
3196         (build_index_type): When we can't hash an index type (e.g.,
3197         because its maximum value is negative), the index type requires
3198         structural equality tests.
3199         (build_array_type): Build the canonical form of an array type.
3200         (build_function_type): Function types require structural equality,
3201         because they contain default arguments, attributes, etc.
3202         (build_method_type_directly): Ditto for method types.
3203         (build_offset_type): Build the canonical offset type.
3204         (build_complex_type): Build the canonical vector type.
3205         (make_vector_type): Build the canonical vector type.
3206         * tree.h (TYPE_CANONICAL): New.
3207         (TYPE_STRUCTURAL_EQUALITY_P): New.
3208         (SET_TYPE_STRUCTURAL_EQUALITY): New.
3209         (struct tree_type): Added "canonical" field.
3210         * params.h (VERIFY_CANONICAL_TYPES): New.
3211         * doc/c-tree.texi (TYPE_CANONICAL): Document.
3212         (TYPE_STRUCTURAL_EQUALITY_P): Document.
3213         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3214         * doc/invoke.texi (verify-canonical-types): Document --param
3215         parameter for verifying canonical types.
3216
3217 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3218
3219         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3220         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3221         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3222         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3223         tm_file.
3224         * config/rs6000/e500.h: New.
3225         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3226         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3227         Remove.
3228         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3229         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3230         TARGET_E500_DOUBLE): Remove.
3231         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3232         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3233         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3234         * config/rs6000/rs6000.c (rs6000_override_options): Use
3235         CHECK_E500_OPTIONS.
3236
3237 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3238
3239         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3240         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3241         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3242         rs6000_emit_cmove): Don't check TARGET_E500.
3243         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3244         sordered): Don't check TARGET_E500.
3245
3246 2007-01-01  Eric Christopher  <echristo@apple.com>
3247
3248         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3249         HARD_REGISTER_NUM_P.
3250
3251 2007-01-01  Roger Sayle  <roger@eyesopen.com>
3252
3253         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3254         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3255         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3256         eq/ne Y".
3257
3258 2007-01-01  Mike Stump  <mrs@apple.com>
3259
3260         * configure.ac: Remove support for building with Apple's gcc-3.1.
3261
3262 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3263
3264         PR middle-end/30311
3265         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3266         REG is a hard register.  Do not modify REG before calling
3267         subreg_nregs.
3268         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3269         multiple of the size of YMODE for certain lowpart cases.
3270
3271 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3272
3273         PR middle-end/30253
3274         * gimplify (voidify_wrapper_expr): Update for
3275         GIMPLE_MODIFY_STMT.
3276
3277 2007-01-01  Andreas Schwab  <schwab@suse.de>
3278
3279         PR target/29166
3280         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3281         save of BR0 in extra_spill_size instead of spill_size.
3282         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3283         area.
3284         (ia64_expand_epilogue): Restore BR0 from its new location.
3285
3286 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3287
3288         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3289         Use a temporary variable if the left hand side is not a gimple
3290         register.
3291
3292 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3293
3294         * gimplify.c (gimplify_return_expr): Make the temporary variable
3295         for the return expression, a gimple register variable.
3296
3297 2007-01-01  Jan Hubicka  <jh@suse.cz>
3298
3299         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3300         INSN_LIST.
3301
3302 2007-01-01  Mike Stump  <mrs@apple.com>
3303
3304         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3305         .literal16.
3306         * config/darwin.c (machopic_select_rtx_section): Use
3307         HAVE_GAS_LITERAL16.
3308         (darwin_mergeable_constant_section): Likewise.
3309         * configure: Regenerate.
3310         * config.in: Regenerate.
3311
3312 2007-01-01  Jan Hubicka  <jh@suse.cz>
3313             Andrew Pinski  <pinskia@gmail.com>
3314
3315         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3316         before starting IPA passes.