OSDN Git Service

2007-01-31 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-02-01  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
2
3         * doc/c-tree.texi (Expression trees): Improve markup.
4         * doc/tm.texi (Register Classes, Addressing Modes)
5         (Floating Point): Fix spacing after abbreviations.  Fix some
6         typos.
7
8 2007-02-01  Ben Elliston  <bje@au.ibm.com>
9
10         * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
11         * doc/contrib.texi: Likewise.
12         * doc/install.texi: Likewise.
13
14 2007-01-31  Richard Henderson  <rth@redhat.com>
15             Ian Lance Taylor  <iant@google.com>
16
17         * lower-subreg.c: New file.
18         * rtl.def (CONCATN): Define.
19         * passes.c (init_optimization_passes): Add pass_lower_subreg and
20         pass_lower_subreg2.
21         * emit-rtl.c (update_reg_offset): New static function, broken out
22         of gen_rtx_REG_offset.
23         (gen_rtx_REG_offset): Call update_reg_offset.
24         (gen_reg_rtx_offset): New function.
25         * regclass.c: Revert patch of 2006-03-05, restoring
26         reg_scan_update.
27         (clear_reg_info_regno): New function.
28         * dwarf2out.c (concatn_loc_descriptor): New static function.
29         (loc_descriptor): Handle CONCATN.
30         * common.opt (fsplit_wide_types): New option.
31         * opts.c (decode_options): Set flag_split_wide_types when
32         optimizing.
33         * timevar.def (TV_LOWER_SUBREG): Define.
34         * rtl.h (gen_reg_rtx_offset): Declare.
35         (reg_scan_update): Declare.
36         * regs.h (clear_reg_info_regno): Declare.
37         * tree-pass.h (pass_lower_subreg): Declare.
38         (pass_lower_subreg2): Declare.
39         * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
40         (Optimize Options): Add -fsplit-wide-types to -O1 list.  Document
41         -fsplit-wide-types.
42         * doc/rtl.texi (Regs and Memory): Document concat and concatn.
43         * Makefile.in (OBJS-common): Add lower-subreg.o.
44         (lower-subreg.o): New target.
45
46 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
47
48         * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
49
50 2007-01-31  Anatoly Sokolov <aesok@post.ru>
51
52         PR target/19087
53         * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
54
55 2007-01-31  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
56
57         PR middle-end/29335
58         * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
59
60 2007-01-31  Zdenek Dvorak <dvorakz@suse.cz>
61
62         * cfgloop.h: Include vec-prim.h.
63         (enum li_flags): Remove LI_ONLY_OLD.
64         (loop_iterator): Changed.
65         (fel_next, fel_init): Iterate over loop tree.
66         (FOR_EACH_LOOP_BREAK): New macro.
67         * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
68         FOR_EACH_LOOP.
69         * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
70         * modulo-sched.c (sms_schedule): Ditto.
71         * tree-vectorizer.c (vectorize_loops): Ditto.
72         * doc/loop.texi: Update information on loop numbering and behavior of
73         FOR_EACH_LOOP wrto new loops.
74         * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
75         add_to_evolution_1): Test nestedness of loops instead of comparing
76         their numbers.
77         * tree-chrec.c (chrec_fold_plus_poly_poly,
78         chrec_fold_multiply_poly_poly, chrec_evaluate,
79         hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
80         reset_evolution_in_loop): Ditto.
81         * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
82
83 2007-01-31  Dirk Mueller  <dmueller@suse.de>
84
85         * c-common.c (warn_about_parentheses): Separate warning about
86         un-parenthized sequence of comparison operators from the one
87         which is supposed to warn about x <= y <= z.
88
89 2007-01-31  Uros Bizjak  <ubizjak@gmail.com>
90
91         * optabs.h (enum optab_index): Add new OTI_isinf.
92         (isinf_optab): Define corresponding macro.
93         * optabs.c (init_optabs): Initialize isinf_optab.
94         * genopinit.c (optabs): Implement isinf_optab using isinf?f2
95         patterns.
96         * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
97         (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
98         using isinf_optab.
99         (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
100         expand_builtin_interclass_mathfn.
101         * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
102         * config/i386/i386.md (UNSPEC_FXAM): New constant.
103         (fxam<mode>2_i387): New insn pattern.
104         (isinf<mode>2) New expander to implement isinf, isinff and isinfl
105         built-in functions as x87 inline asm.
106
107 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
108
109         * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
110
111 2007-01-30  Eric Christopher  <echristo@apple.com>
112
113         * config.gcc: Add geode.
114
115 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
116
117         * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
118         ipa-inline.c, tree-profile.c, tree-ssa-live.c,
119         tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
120         value-prof.c: Fix comment typos.
121
122 2007-01-30  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
123
124         PR c++/24745
125         * doc/invoke.texi (Wpointer-arith): Document warning.
126
127 2007-01-30  Janis Johnson
128
129         * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
130         the list of discrepancies from the draft TR.
131
132 2007-01-30  Dirk Mueller  <dmueller@suse.de>
133
134         PR c++/30601
135         * doc/invoke.texi (-Wreturn-type): Update description to
136         match new behavior.
137
138 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
139
140         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
141
142 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
143
144         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
145         to handle optabs that operate on floating point input argument and
146         output to integer output.
147         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
148         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
149         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
150         expand_builtin_interclass_mathfn ().
151         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
152         (ilogbsi2): Remove.
153         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
154         ilogbl built-in functions as x87 intrinsics.
155
156 2007-01-30  Richard Guenther  <rguenther@suse.de>
157
158         PR middle-end/27657
159         * dwarf2out.c (reference_to_unused): Query varpool if the
160         variable was output.
161
162 2007-01-30  Richard Guenther  <rguenther@suse.de>
163
164         PR middle-end/30313
165         * passes.c (execute_one_pass): Reset in_gimple_form to not
166         confuse non-unit-at-a-time mode.
167
168 2007-01-29  Roger Sayle  <roger@eyesopen.com>
169             Richard Guenther  <rguenther@suse.de>
170
171         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
172         avoid undefined behaviour on overflow.  Use force_fit_type_double
173         to construct the constant with the specified TREE_OVERFLOW.
174
175 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
176
177         * config/dfp-bit.c: Add parameterized support for fp exceptions.
178         * config/dfp-bit.h: Ditto.
179
180 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
181
182         * c-decl.c (pop_scope): Replace warnings with call to
183         warn_for_unused_label.
184         * c-common.h (warn_for_unused_label): Declare.
185         * c-common.c (warn_for_unused_label): Define.
186
187 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
188
189         * tree-optimize.c (update_inlined_to_pointers): Delete unused
190         function.
191
192 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
193
194         * Makefile.in (USER_H): Remove decfloat.h.
195         * ginclude/decfloat.h: Delete, moving contents to ...
196         * ginclude/float.h: Add support for decimal floating point,
197         guarded by __STDC_WANT_DEC_FP__.
198
199 2007-01-29  Mike Stump  <mrs@apple.com>
200
201         * doc/gccint.texi (Top): Rename Loop Representation to Loop
202         Analysis and Representation to resolve case insensitive conflict.
203         * doc/loop.texi (Loop Analysis and Representation): Likewise.
204
205 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
206
207         * tree.h (struct tree_memory_tag): Add aliases member.
208         (MTAG_ALIASES): New macro.
209         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
210         (add_may_alias): Remove pointer-set. Update for may_aliases being
211         a bitmap.
212         (mark_aliases_call_clobbered): Update for may_aliases being a
213         bitmap.
214         (compute_tag_properties): Ditto.
215         (create_partition_for): Ditto.
216         (compute_memory_partitions): Ditto.
217         (dump_may_aliases_for): Ditto.
218         (is_aliased_with): Ditto.
219         (add_may_alias_for_new_tag): Ditto.
220         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
221         (compute_is_aliased): New function.
222         (compute_may_aliases): Call compute_is_aliased).
223         (init_alias_info): Initialize alias_bitmap_obstack.
224         (union_alias_set_into): New function.
225         (compute_flow_sensitive_aliasing): Use union_aliases_into.
226         (have_common_aliases_p): Rewrite to take two bitmaps and use
227         intersection.
228         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
229         Update for bitmaps.
230         (finalize_ref_all_pointers): Update for add_may_alias changes.
231         (new_type_alias): Ditto.
232         * tree-flow-inline.h (may_aliases): Return a bitmap.
233         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
234         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
235         may_aliases being a bitmap.
236         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
237         may_aliases now returns a bitmap.
238         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
239         being a bitmap.
240         * tree-ssa-operands.c (add_virtual_operand): Update for
241         may_aliases being a bitmap.
242
243 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
244
245         PR tree-optimization/30630
246         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
247         variable as changed.
248
249 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
250
251         PR c++/28266
252         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
253         expanded only once even if an error occurs.
254
255 2007-01-29  Ben Elliston  <bje@au.ibm.com>
256
257         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
258         (__gcov_average_profiler): Likewise.
259         (__gcov_ior_profiler): Likewise.
260         (__gcov_merge_ior): Likewise.
261
262 2007-01-28  Jan Hubicka  <jh@suse.cz>
263
264         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
265         * value-prof.c (stringop_block_profile): Likewise.
266
267 2007-01-28  Jan Hubicka  <jh@suse.cz>
268
269         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
270         variant handling histograms; add wrapper.
271         (clear_storage_via_libcall): Export.
272         (emit_block_move_hints): Break out from ...; add histograms.
273         (emit_block_move): ... this one.
274         (clear_storage_hints): Break out from ...; add histograms.
275         (clear_storage): ... this one.
276         (set_storage_via_memset): Handle histogram.
277         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
278         (clear_storage_hints, clear_storage_via_libcall): Declare.
279         (set_storage_via_setmem): Update prototype.
280         * doc/md.texi (movmem, setmem): Document new arguments.
281
282         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
283         new histograms.
284         (stringop_block_profile): New global function.
285         (tree_stringops_values_to_profile): Profile block size and alignment.
286         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
287         HIST_TYPE_IOR.
288         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
289         (stringop_block_profile): Declare.
290         * builtins.c: Include value-prof.h.
291         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
292         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
293         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
294         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
295         * profile.c (instrument_values): Add new counters.
296         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
297         calls.
298         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
299         (tree_init_edge_profiler): Build new profilers.
300         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
301         (pass_tree_profile): Add dump.
302         (tree_profile_hooks): Update.
303         * Makefile.in (LIBGCOV): Add new constants.
304         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
305         __gcov_ior_profiler): New.
306         * i386.md (movmem/setmem expanders): Add new optional arguments.
307
308 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
309
310         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
311
312 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
313
314         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
315         (tree_unswitch_single_loop): Pass eni_size_weights to
316         tree_num_loop_insns.
317         * tree-ssa-loop-manip.c: Include tree-inline.h.
318         (can_unroll_loop_p): Pass eni_size_weights to
319         tree_num_loop_insns.
320         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
321         Pass eni_size_weights to estimate_num_insns.
322         * tree.h (init_inline_once): Export.
323         * toplev.c (backend_init): Call init_inline_once.
324         * cgraphunit.c (cgraph_process_new_functions,
325         cgraph_analyze_function): Pass eni_inlining_weights to
326         estimate_num_insns.
327         * ipa-inline.c (compute_inline_parameters): Ditto.
328         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
329         to estimate_num_insns.
330         (try_unroll_loop_completely): Pass eni_size_weights to
331         tree_num_loop_insns.
332         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
333         ot estimate_num_insns.
334         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
335         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
336         * tree-inline.c (eni_inlining_weights, eni_size_weights,
337         eni_time_weights): New variables.
338         (init_inline_once): Initialize them.
339         (struct eni_data): Mew.
340         (estimate_num_insns_1, estimate_num_insns): Use weights.
341         * tree-inline.h (struct eni_weights_d): New.
342         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
343         (estimate_num_insns): Declaration changed.
344         * cfgloop.h (tree_num_loop_insns): Declaration changed.
345         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
346         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
347
348 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
349
350         * tree-data-ref.c (conflict_fn): Assert that the number of affine
351         relations in the conflict function is valid.
352
353 2007-01-27  Ian Lance Taylor  <iant@google.com>
354
355         * common.opt: Add fstrict-overflow.
356         * opts.c (decode_options): Set flag_strict_overflow if -O2.
357         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
358         (TYPE_OVERFLOW_UNDEFINED): Define.
359         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
360         Replace all uses.
361         * tree.h (TYPE_TRAP_SIGNED): Don't define.
362         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
363         (fold_negate_expr): Likewise.
364         (make_range): Likewise.
365         (extract_muldiv_1): Likewise.
366         (maybe_canonicalize_comparison): Likewise.
367         (fold_comparison): Likewise.
368         (fold_binary): Likewise.
369         (tree_expr_nonnegative_p): Likewise.
370         (tree_expr_nonzero_p): Likewise.
371         * tree-vrp.c (compare_values): Likewise.
372         (extract_range_from_binary_expr): Likewise.
373         (extract_range_from_unary_expr): Likewise.
374         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
375         Likewise.
376         (nowrap_type_p): Likewise.
377         * tree-scalar-evolution.c (simple_iv): Likewise.
378         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
379         (build_range_check): Likewise.
380         (extract_muldiv_1): Likewise.
381         (fold_comparison): Likewise.
382         * tree-vrp.c (vrp_int_const_binop): Likewise.
383         (extract_range_from_unary_expr): Likewise.
384         * convert.c (convert_to_integer): Likewise.
385         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
386         (fold_comparison): Likewise.
387         (fold_binary): Likewise.
388         * optabs.c (optab_for_tree_code): Likewise.
389         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
390         * simplify-rtx.c (simplify_const_relational_operation): Check
391         flag_strict_overflow and flag_trapv.
392         (simplify_const_relational_operation): Likewise.
393         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
394         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
395         -fstrict-overflow.
396
397 2007-01-27  Roger Sayle  <roger@eyesopen.com>
398
399         * tree.c (tree_fold_gcd): Delete.
400         * tree.h (tree_fold_gcd): Remove prototype.
401         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
402         test whether one constant integer is a multiple of another.  Instead
403         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
404         * fold-const.c (multiple_of_p):  We've determined both TOP and
405         BOTTOM are integer constants so we can call int_const_binop directly
406         instead of the more generic const_binop.
407
408 2007-01-27  Roger Sayle  <roger@eyesopen.com>
409
410         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
411         1*X check that the constant hasn't overflowed, to preserve the
412         TREE_OVERFLOW bit.
413         (round_up): Provide an efficient implementation when rouding-up an
414         INTEGER_CST to a power-of-two.
415
416 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
417
418         * doc/sourcebuild.texi: Add comma for clarity.
419         * doc/extend.texi: Fix some typos.
420         * doc/passes.texi: Likewise.
421         * doc/cppinternals.texi: Likewise.
422         * doc/c-tree.texi: Likewise.
423         * doc/tree-ssa.texi: Likewise.
424         * doc/install.texi: Likewise.
425
426 2007-01-27  Jan Hubicka  <jh@suse.cz>
427
428         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
429         (pass_sra): Do not require alias information.
430         * passes.c (init_optimization_passes): Add SRA
431
432 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
433
434         * tracer.c (rest_of_handle_tracer): We already cleaned
435         up the CFG in tracer() so don't do it here again.
436         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
437         cleanup_cfg here, either.  And don't call renumber_insns.
438
439         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
440         and BB_HEAD are in the insn stream here.  Instead make sure
441         that BB_INSN is valid on all insns.  Also, do check here that
442         there are no pending branch predictions...
443         (rtl_verify_flow_info): ...instead of doing it here.  Checks
444         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
445         here.
446
447 2007-01-26  Roger Sayle  <roger@eyesopen.com>
448
449         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
450         function to simplify/factorize operand order canonicalization.
451         (ix86_fixup_binary_operands): Reorganize using the above function.
452         (ix86_binary_operator_ok): Likewise.
453
454 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
455
456         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
457         file scope from optimize_attrs.
458         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
459         up search.
460         (optimize_attrs): Clear insn_code_values after freeing it.
461
462 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
463
464         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
465         a parts.base assignment.
466
467 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
468
469         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
470         (compute_subscript_distance, initialize_data_dependence_relation,
471         finalize_ddr_dependent, analyze_ziv_subscript,
472         analyze_siv_subscript_cst_affine,
473         compute_overlap_steps_for_affine_univar,
474         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
475         analyze_siv_subscript, analyze_miv_subscript,
476         analyze_overlapping_iterations, subscript_dependence_tester_1,
477         compute_self_dependence, free_dependence_relation): Work
478         with affine_fn instead of chrecs.
479         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
480         common_affine_function, affine_function_base,
481         affine_function_constant_p, affine_fn_op, affine_fn_plus,
482         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
483         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
484         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
485         (all_chrecs_equal_p): Removed.
486         * tree-data-ref.h (affine_fn, conflict_function): New types.
487         (struct subscript): Change type of conflicting_iterations_in_a
488         and conflicting_iterations_in_b.
489
490 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
491
492         PR other/30182
493         * config/pa/pa.h (TARGET_HPUX_11): New.
494         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
495         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
496
497 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
498         Richard Guenther  <rguenther@suse.de>
499
500         * tree-ssa-structalias.c (solve_graph): Handle case
501         we merged the variable to another.
502
503 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
504
505         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
506         functions.
507
508         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
509         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
510
511 2007-01-25  DJ Delorie  <dj@redhat.com>
512
513         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
514         allow changes to modes which don't fit in those registers.
515
516         * reload1.c (choose_reload_regs): Check for invalid subregs before
517         computing their locations, not after.
518
519 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
520
521         PR 25127
522         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
523         save Altivec registers in an eh_return function.
524         (compute_vrsave_mask): Likewise.
525         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
526         for saving Altivec registers.
527         (rs6000_emit_prologue): Don't allocate stack twice in
528         eh_return function.  Correct expected value of altivec_save_offset
529         when using save_world.  Describe save of R0 to stack when using
530         save_world.  Describe stack pointer adjustment when using
531         save_world.  Remove duplicated eh_return parameter register saving.
532         Update sp_offset variable after save_world.
533         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
534         darwin-world.asm.
535         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
536         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
537         implies Altivec.
538
539 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
540
541         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
542
543 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
544
545         PR other/30182
546         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
547         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
548
549 2007-01-25  Richard Guenther  <rguenther@suse.de>
550
551         * doc/invoke.texi (-Wcoverage-mismatch): Document.
552         * common.opt (-Wcoverage-mismatch): New warning option.
553         * coverage.c (get_coverage_counts): Ignore coverage mismatch
554         if -Wcoverage-mismatch is given.
555
556 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
557
558         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
559         (ipcp_driver): Change to static definition.
560         Add dumping of the ifunctions.
561         (constant_val_insert): Remove unused parameter. Support for SSA.
562         (ipcp_propagate_const): Support for SSA.
563         (ipcp_profile_bb_print): Print only analyzed nodes.
564         (ipcp_replace_map_create): Remove support for Fortran constant
565         for now.
566         * ipa-prop.c (ipa_method_modify_stmt,
567         ipa_callsite_compute_param): Support for SSA.
568         * ipa-prop.h (ipcp_driver): Remove declaration.
569         (IS_VALID_TREE_MAP_INDEX): Add define.
570
571 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
572
573         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
574
575 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
576
577         PR middle-end/30447
578         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
579         constant argument to cabs and do it without checking for
580         -funsafe-math-optimizations.
581
582 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
583
584         * c-common.h (RID_FIRST_CXX0X): New.
585         (RID_LAST_CXX0X): New.
586         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
587         by -Wall.
588         * c.opt (Wc++0x-compat): New.
589         * doc/invoke.texi (-Wc++0x-compat): Document.
590
591 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
592
593         * builtins.c (fold_builtin_carg): New.
594         (fold_builtin_1): Use it.
595
596 2007-01-24  Jan Hubicka  <jh@suse.cz>
597
598         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
599
600         * ipa-inline.c (initial_insns, max_insns): Delete.
601         (compute_max_insns): New function.
602         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
603         of insns as base for code growth.
604         (cgraph_decide_inlining): Make initial_insns local; do not compute
605         max_insns.
606         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
607         * doc/invoke.texi (inline-unit-growth): Update docs.
608
609 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
610
611         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
612         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
613         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
614         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
615         for CPUs that have PTA_CX16 set.
616
617         PR middle-end/27416
618         * gimplify.c (omp_check_private): New function.
619         (gimplify_scan_omp_clauses): Use it for
620         firstprivate/lastprivate/reduction.
621
622         PR middle-end/30494
623         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
624         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
625
626         PR middle-end/30421
627         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
628         If lower_lastprivate_clauses emits some statements, append them
629         to dlist rather than body_p and to body_p append an initializer.
630         (lower_omp_for): Adjust caller.
631
632 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
633
634         * target.h (globalize_decl_name): New.
635         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
636         * output.h (default_globalize_decl_name): New.
637         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
638         globalize_label.
639         (globalize_decl): Ditto.
640         (default_globalize_decl_name): New.
641         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
642         (ia64_handle_version_id_attribute): New.
643         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
644         (ia64_asm_output_external): Use globalize_decl_name instead
645         of globalize_label.
646         * doc/extend.texi (version_id): New pragma.
647         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
648
649 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
650
651         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
652         _uleb128_t and _Unwind_SWord with _sleb128_t.
653         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
654         uw_frame_state_for, uw_update_context_1): Likewise.
655         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
656         * unwind-pe.h (read_uleb128, read_sleb128,
657         read_encoded_value_with_base): Likewise.
658         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
659
660 2007-01-24  Richard Guenther  <rguenther@suse.de>
661
662         * builtins.c (expand_builtin_cexpi): Get the fndecl
663         for cexp in the correct way.
664
665 2007-01-24  Jan Hubicka  <jh@suse.cz>
666
667         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
668
669 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
670
671         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
672         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
673         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
674         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
675         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
676         __SIZEOF_PTRDIFF_T__.
677         * doc/cpp.texi: Documentation for the new builtin macros added.
678
679 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
680
681         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
682         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
683         instead of operands[2] to avoid access past the end of array.
684
685 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
686
687         * reload1.c (emit_reload_insns): Pass the reload register
688         for a non-spill output reload through forget_old_reloads_1.
689
690 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
691
692         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
693         _SOFT_DOUBLE if doubles use software floating-point.
694         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
695         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
696         * config/rs6000/darwin-ldouble.c: Also compile functions for
697         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
698         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
699         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
700         commented-out long double override.
701         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
702         * config/rs6000/eabispe.h: Likewise.
703         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
704         long double for non-SPE.
705         (rs6000_handle_option): Likewise.
706         (invalid_e500_subreg): Disallow more subregs involding DImode,
707         DFmode, TImode or TFmode.
708         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
709         E500 double.
710         (legitimate_lo_sum_address_p): Also check for TFmode for E500
711         double.
712         (rs6000_legitimize_address): Also handle TFmode for E500 double.
713         (rs6000_legitimize_reload_address): Also handle TFmode for E500
714         double.
715         (rs6000_legitimate_address): Also check for TFmode for E500
716         double.
717         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
718         (spe_build_register_parallel): Handle TFmode and TCmode.
719         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
720         double.
721         (function_arg): Handle TFmode and TCmode for E500 double.
722         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
723         in general.
724         (print_operand): Handle TFmode and TImode for %y.
725         (rs6000_generate_compare): Handle TFmode comparisons for E500
726         double.
727         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
728         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
729         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
730         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
731         TFmode for E500 double.
732         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
733         (floatsidf2): Enable for E500 double.
734         (movtf_softfloat): Use rs6000_nonimmediate_operand.
735         (extenddftf2): Change to extenddftf2_fprs.
736         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
737         depending on TARGET_E500_DOUBLE.
738         (extendsftf2): Enable for E500 double.
739         (trunctfdf2): Enable for E500 double.
740         (trunctfsf2): Change to trunctfsf2_fprs.
741         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
742         depending on TARGET_E500_DOUBLE.
743         (floatsitf2): Enable for E500 double.
744         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
745         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
746         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
747         (negtf2): Change to negtf2_internal.
748         (negtf2): New expander.
749         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
750         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
751         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
752         (movdi_internal32): Use rs6000_nonimmediate_operand.
753         (unnamed splitter): Likewise.
754         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
755         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
756         (SPE64TF, DITI): New mode macros.
757         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
758         modes.
759         (frob_tf_ti): New.
760         (frob_<mode>_di_2): New.
761         (frob_tf_di_8_2): New.
762         (frob_di_df): Change to frob_di_<mode>; allow more modes.
763         (frob_ti_tf): New.
764         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
765         more modes.
766         (frob_ti_<mode>_8_2): New.
767         (frob_ti_tf_2): New.
768         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
769         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
770         TFmode.
771         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
772         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
773         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
774         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
775         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
776         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
777         tsttflt_gp): New.
778
779 2007-01-23  Ian Lance Taylor  <iant@google.com>
780
781         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
782         insn-*.o first.
783         (OBJS-archive): Reformat, alphabetize.
784         (OBJS): Change out_object_file to OBJS-md.
785
786 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
787
788         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
789         JUMP_P and CALL_P predicates where applicable.
790         * config/i386/i386.c: Ditto.
791         * config/i386/i386.md: Ditto.
792         * config/i386/mmx.md: Ditto.
793         * config/i386/predicates.md: Ditto.
794
795 2007-01-22  Andreas Schwab  <schwab@suse.de>
796
797         * config/m68k/m68k.h: Fix comment.
798
799 2007-01-22  Jan Hubicka  <jh@suse.cz>
800
801         * passes.c (init_optimization_passes): Do not rerun
802         pass_early_warn_uninitialized.
803
804 2007-01-22  Richard Guenther  <rguenther@suse.de>
805
806         PR tree-optimization/30038
807         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
808         function.
809         (execute_cse_sincos_1): Likewise.
810         (execute_cse_sincos): Likewise.
811         (gate_cse_sincos): Likewise.
812         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
813         the cexpi() canonicalization of sincos().
814         * tree-pass.h (pass_cse_sincos): Declare.
815         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
816
817 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
818
819         PR rtl-optimization/29329
820         * combine.c (replaced_rhs_insn): Rename to i2mod.
821         (replaced_rhs_value): Rename to i2mod_new_rhs.
822         (i2mod_old_rhs): New global variable.
823         (combine_instructions): Adjust for above change.  Save a copy of
824         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
825         note are substituted in the second instruction.
826         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
827         ditch the note if it pertains to the second eliminated register
828         and this register is mentioned in i2mod_old_rhs.
829
830         Revert:
831         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
832
833         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
834         the insn to which the note was originally attached.
835
836 2007-01-21  Jan Hubicka  <jh@suse.cz>
837
838         * ipa-inline.c (inlining_mode): Comment, move up.
839         (cgraph_decide_inlining_incrementally): Do not perform inlining
840         itself; fix handling of flattening of self recursive functions.
841         (cgraph_find_cycles): Remove.
842         (cgraph_flatten_node): Remove.
843         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
844         (try_inline): New function.
845         (cgraph_early_inlining): Update call of
846         cgraph_decide_inlining_incrementally.  Apply inlining here.
847         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
848
849 2007-01-21  Dirk Mueller  <dmueller@suse.de>
850
851         PR bootstrap/30511
852         * tree-vrp.c (check_array_bounds): do not warn
853         about ADDR_EXPR's of ARRAY_REF's which are immediately
854         used in binary expressions.
855
856 2007-01-21  Ira Rosen  <irar@il.ibm.com>
857
858         * tree-vectorizer.h (struct _stmt_vec_info): Add new field
859         read_write_dep and macros for its access.
860         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
861         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
862         argument, call vect_check_interleaving for every independent pair of
863         data-refs. Mark loads that access the same memory location as a store
864         in the loop.
865         (vect_check_dependences): Remove.
866         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences
867         call, fix the call to vect_analyze_data_ref_dependence.
868         (vect_analyze_data_ref_access): For statements that access the same
869         data-ref, check that they are not stores; for loads, check that there
870         is no store that access the same location.
871
872 2007-01-20  Roger Sayle  <roger@eyesopen.com>
873             Joseph Myers  <joseph@codesourcery.com>
874
875         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
876         (-fdump-rtl-stack): Likewise.
877         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
878         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
879         (-mpt-fixed): Use @minus{} for minus sign.
880         (Using Precompiled Headers): Fix under/overfull hbox.
881
882 2007-01-20  Jan Hubicka  <jh@suse.cz>
883
884         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
885         place within annotation.
886
887 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
888
889         * pa.c (output_move_double): Change array size of xoperands to 4.
890
891 2007-01-20  Jan Hubicka  <jh@suse.cz>
892
893         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
894         as temporary.
895
896 2007-01-19  Ian Lance Taylor  <iant@google.com>
897
898         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
899         constant to unsigned HOST_WIDE_INT before negating.
900
901 2007-01-19  Ian Lance Taylor  <iant@google.com>
902
903         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
904         to unsigned.
905         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
906         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
907         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
908         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
909         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
910         (realloc_vdef, realloc_vuse): Change second parameter to
911         unsigned.
912         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
913         fields to unsigned.
914         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
915         unsigned. Change x and lim locals to unsigned.
916         (realloc_vdef, realloc_vuse): Change num_elem parameter to
917         unsigned.
918         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
919         (copy_virtual_operands): Change i and n locals to unsigned.
920
921 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
922
923         PR middle-end/29335
924         * builtins.c (fold_builtin_1): Handle builtin fdim.
925
926 2007-01-20  Jan Hubicka  <jh@suse.cz>
927
928         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
929         * tree-ssa-structalias.c (compute_points_to_sets): Do call
930         init_alias_heapvars.
931         (init_alias_heapvars): Initialize only when not already initialized.
932         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
933
934 2007-01-19  Roger Sayle  <roger@eyesopen.com>
935
936         * common.opt (fsigned-zeros): New command line option.
937         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
938         of flag_unsafe_math_optimizations.
939         * opts.c (set_fast_math_flags): The -ffast-math command line option
940         implies -fno-signed-zeros.
941         (fast_math_flags_set_p): Likewise.
942
943         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
944         the documentation of -ffast-math appropriately.  Wrap long lines.
945
946 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
947
948         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
949         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
950
951 2007-01-19  Tomas Bily  <tbily@suse.cz>
952
953         * cgraphunit.c (cgraph_finalize_function): Updating of pid
954         * tree-profile.c:
955         (tree_init_ic_make_global_vars): New function
956         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
957         (tree_gen_ic_profiler): New function
958         (tree_gen_ic_func_profiler): New function
959         (tree_profiling): Added calling of tree_gen_ic_func_profiler
960         (tree_profile_hooks): Added hook for indirec/virtual calls
961         * value-prof.c (tree_find_values_to_profile): New case for
962         indirect calls
963         (tree_values_to_profile): Call for determining indirect/virtual
964         counters
965         (tree_indirect_call_to_profile): New function
966         (tree_ic_transform): New function
967         (tree_ic): New function
968         (find_func_by_pid): New function
969         (init_pid_map): New function
970         (tree_value_profile_transformations): Added check for
971         indirect/virtual call transformation
972         * value-prof.h (enum hist_type): New counter type for
973         indirect/virtual calls
974         (profile_hooks): Added new hook for profiling indirect/virtual
975         calls
976         * profile.c (instrument_values): New case for indirect/virtual
977         call added
978         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
979         (GCOV_COUNTER_V_INDIR): New counter type
980         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
981         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
982         call added
983         * cgraph.c: Definition of cgraph_max_pid
984         (cgraph_create_node): Default init of pid attribute
985         * cgraph.h: Declaration of cgraph_max_pid
986         (struct cgraph_node): Added pid attribute
987         * libgcov.c (__gcov_indirect_call_profiler): New function
988         (__gcov_one_value_profiler_body): New function
989         (__gcov_one_value_profiler): Body was moved to
990         __gcov_one_value_profiler_body and calls it
991
992 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
993
994         * doc/gty.texi (Options): Document the mark_hook option to GTY.
995         * gengtype.c (write_types_data, write_func_for_structure,
996         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
997         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
998         (walk_type, write_func_for_structure): Generate the mark_hook if
999         needed.
1000
1001 2007-01-19  Jan Hubicka  <jh@suse.cz>
1002
1003         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1004         'early' argument take inlining mode argument specifying whether to
1005         inline for size/speeed or all functions; add support for flattening;
1006         improve dumpting.
1007         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1008
1009 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1010
1011         PR c++/17947
1012         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1013         the name of the declared identifier.
1014
1015 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1016
1017         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1018         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1019         duplicate evaluation.
1020
1021 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1022
1023         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1024         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
1025         to XFMode, use acosxf2 and truncate result to requested mode.
1026         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1027         (asin<mode>2): Similarly, with asin expanders.
1028         (*fscalexf4_i387): Rename from *fscalexf4.
1029         (expNcorexf3): New expander.
1030         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1031         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1032         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
1033         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1034         patterns for SSE math.
1035         (exp10<mode>2): Similarly, with exp10 expanders.
1036         (exp2<mode>2): Similarly, with exp2 expanders.
1037         (expm1<mode>2): Similarly, with expm1 expanders.
1038         (ldexp<mode>3): Similarly, with ldexp expanders.
1039         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1040         Use gen_truncxf<mode>2_i387_noop to truncate result.
1041
1042 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1043
1044         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1045         for ColdFire targets.
1046
1047 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
1048             Richard Sandiford  <richard@codesourcery.com>
1049
1050         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1051         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1052         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1053         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1054         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1055         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1056
1057 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1058
1059         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1060         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
1061         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
1062         (*adddi_dilshr32_cf): New pattern.
1063         (adddi3, subdi3): Remove first alternatives.
1064
1065 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1066
1067         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1068         implemented using move.l, do not use its cc result for floating-point
1069         comparisons.
1070
1071 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1072
1073         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1074         from accepting 's' constraints if flag_pic.
1075
1076 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1077
1078         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1079         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1080         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1081         condition from TARGET_68881 to TARGET_HARD_FLOAT.
1082
1083 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
1084
1085         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1086         ColdFire alternatives.
1087         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1088         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
1089         architectures that support ff1 instruction.
1090
1091 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1092             Julian Brown  <julian@codesourcery.com>
1093
1094         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1095         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1096         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
1097         Add commentary.
1098
1099 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1100
1101         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1102         and return a bool.
1103         (output_move_const_into_data_reg, output_move_simode_const): Delete.
1104         * config/m68k/m68k.c (const_method, const_int_cost): Take a
1105         HOST_WIDE_INT instead of an rtx.
1106         (m68k_rtx_costs): Update call accordingly.
1107         (output_move_const_into_data_reg): Likewise.  Fix formatting.
1108         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1109         Return a bool.
1110         (output_move_simode_const): Update calls after above changes.
1111         Rework to use automatic variables and predicates like MEM_P.
1112         * config/m68k/m68k.md (pushexthisi_const): Update call to
1113         valid_mov3q_const.
1114
1115 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1116
1117         * tree-ssa-alias.c (perform_var_substitution): Fix typo
1118         in dump_flags test.
1119
1120 2007-01-19  Richard Guenther  <rguenther@suse.de>
1121
1122         * builtins.c (expand_builtin_cexpi): Fall back to expanding
1123         via cexp in case sincos is not available.
1124
1125 2007-01-19  Richard Guenther  <rguenther@suse.de>
1126
1127         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1128         * defaults.h (TARGET_HAS_SINCOS): Default to off.
1129         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1130         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1131         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1132         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1133         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1134         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1135
1136 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1137
1138         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1139         New insn patterns.
1140         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1141         (atan2xf3): Directly generate RTL pattern.
1142         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1143         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1144         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1145         disable patterns for SSE math.
1146         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1147         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1148         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1149         disable patterns for SSE math.
1150
1151 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1152
1153         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1154
1155 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1156
1157         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1158         VIEW_CONVERT_EXPR to the same type as it's operand.
1159
1160 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1161
1162         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1163         configured for long double 128.
1164
1165 2007-01-18  Mike Stump  <mrs@apple.com>
1166
1167         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1168         error.
1169
1170 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1171
1172         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1173         type correct.
1174
1175 2007-01-18  Jan Hubicka  <jh@suse.cz>
1176
1177         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1178         greater than NUM_VOP_FREE_BUCKETS.
1179
1180 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1181
1182         * tree-ssa-structalias.c: Update comments.
1183         (ptabitmap_obstack): Removed.
1184         (pta_obstack): New.
1185         (oldpta_obstack): Ditto.
1186         (stats): Add a few members.
1187         (struct variable_info): Remove node, complex, address_taken, and
1188         indirect_target members. Add oldsolution member.
1189         (new_var_info): Do not initialize removed members.
1190         (constraint_expr_type): Remove INCLUDES.
1191         (constraint_graph): Add size, implicit_preds, rep,
1192         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1193         (FIRST_REF_NODE): New macro.
1194         (LAST_REF_NODE): Ditto.
1195         (FIRST_ADDR_NODE): Ditto.
1196         (find): New function.
1197         (unite): Ditto.
1198         (dump_constraint): Do not handle INCLUDES.
1199         (insert_into_complex): Do not insert duplicate constraints.
1200         (condense_varmap_nodes): Renamed and rewritten into ...
1201         (merge_node_constraints): This. Also fix bug in handling of
1202         offseted copy constraints.
1203         (clear_edges_for_node): No longer need to deal with preds at all,
1204         or removing associated preds/succs.
1205         (merge_graph_nodes): Deal with indirect_cycles.
1206         Don't deal with predecessors.
1207         (add_implicit_graph_edge): New function.
1208         (add_pred_graph_edge): Ditto.
1209         (add_graph_edge): Don't deal with predecessors.
1210         (build_constraint_graph): Removed.
1211         (build_pred_graph): New function.
1212         (build_succ_graph): Ditto.
1213         (struct scc_info): Removed in_component. Added roots, dfs, and
1214         node_mapping. Remove visited_index, unification_queue.
1215         (scc_visit): Deal with union-find we do now.
1216         Deal with cycles with REF nodes.
1217         (collapse_nodes): Renamed and rewritten to ...
1218         (unify_nodes): This.
1219         (process_unification_queue): Removed.
1220         (topo_visit): Cleanup
1221         (do_da_constraint): Use find.
1222         (do_sd_constraint): Ditto.
1223         (do_ds_constraint): Ditto.
1224         (do_complex_constraint): Ditto.
1225         (init_scc_info): Update for removed and added members.
1226         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1227         (find_indirect_cycles): This.
1228         (equivalence_class): New variable.
1229         (label_visit): New function.
1230         (perform_variable_substitution): Rewritten.
1231         (free_var_substitution_info): New function.
1232         (find_equivalent_node): Ditto.
1233         (move_complex_constraints): Ditto.
1234         (eliminate_indirect_cycles): Ditto.
1235         (solve_graph): Only propagate changed bits.
1236         Use indirect cycle elimination.
1237         Use find.
1238         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1239         (tree_id_eq): Renamed to ...
1240         (tree_vi_eq): This. Update for member change
1241         (insert_id_for_tree): Renamed and rewritten to ...
1242         (insert_vi_for_tree): This.
1243         (lookup_id_for_tree): Renamed and rewritten to ...
1244         (lookup_vi_for_tree): This.
1245         (get_id_for_tree): Renamed and rewritten to ...
1246         (get_vi_for_tree): Ditto.
1247         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1248         (process_constraint): Don't handle INCLUDES.
1249         Remove special ADDRESSOF case.
1250         (find_func_aliases): Rewrite to use vi functions instead of id
1251         ones.
1252         (create_function_info_for): Ditto.
1253         (create_variable_info_for): Ditto.
1254         (intra_create_variable_infos): Ditto.
1255         (merge_smts_into): Ditto.
1256         (find_what_p_points_to): Ditto.
1257         (init_base_vars): Ditto.
1258         (init_alias_vars): Ditto.
1259         (remove_preds_and_fake_succs): New function.
1260         (dump_sa_points_to_info): Dump new stats.
1261         (dump_solution_for_var): Use find.
1262         (set_used_smts): Fix formatting.
1263         (compute_points_to_sets): Updated for new functions.
1264         (ipa_pta_execute): Ditto.
1265
1266 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1267             Richard Sandiford  <richard@codesourcery.com>
1268
1269         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1270         parallels.
1271         * calls.c (expand_call): If the return value is a PARALLEL,
1272         extract its first member.
1273         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1274         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1275         use FUNCTION_EXTRA_EPILOGUE.
1276         (m68k_function_value): Return a PARALLEL if the return value
1277         is of a pointer type.
1278         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1279         (FUNCTION_EXTRA_EPILOGUE): Remove.
1280         * config/m68k/m68k.md (D0_REG): New constant.
1281
1282 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1283
1284         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1285         output a NOP for empty epilogues.
1286
1287 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1288
1289         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1290         before function.  Extend register save check to include all
1291         registers, not just integer ones.
1292
1293 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1294
1295         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1296         (m68k_use_return_insn): ...this.
1297         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1298         * config/m68k/m68k.c (use_return_insn): Rename to...
1299         (m68k_use_return_insn): ...this.
1300         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1301         of USE_RETURN_INSN.
1302
1303 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1304
1305         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1306         nearest even, fix denormal rounding overflow.
1307
1308 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1309
1310         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1311         f%$smove and f%$move.
1312         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1313         (extendsfdf2_cf): Use fdmove instead of f%&move.
1314         (truncdfsf2_cf): Use fsmove instead of f%$smove.
1315         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1316
1317 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1318
1319         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1320         GPR<-GPR moves.
1321
1322 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1323
1324         * real.h (real_format): Add a canonical_nan_lsbs_set field.
1325         (coldfire_single_format): Declare.
1326         (coldfire_double_format): Likewise.
1327         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1328         of qnan_msb_set to determine the lower bits of a canonical
1329         NaN significand.
1330         (encode_ieee_double): Likewise.
1331         (encode_ieee_quad): Likewise.
1332         (ieee_single_format): Initialize canonical_nan_lsbs_set.
1333         (mips_single_format): Likewise.
1334         (ieee_double_format): Likewise.
1335         (mips_double_format): Likewise.
1336         (ieee_extended_motorola_format): Likewise.
1337         (ieee_extended_intel_96_format): Likewise.
1338         (ieee_extended_intel_128_format): Likewise.
1339         (ieee_extended_intel_96_round_53_format): Likewise.
1340         (ibm_extended_format): Likewise.
1341         (mips_extended_format): Likewise.
1342         (ieee_quad_format): Likewise.
1343         (mips_quad_format): Likewise.
1344         (vax_f_format): Likewise.
1345         (vax_d_format): Likewise.
1346         (vax_g_format): Likewise.
1347         (i370_single_format): Likewise.
1348         (i370_double_format): Likewise.
1349         (decimal_single_format): Likewise.
1350         (decimal_double_format): Likewise.
1351         (decimal_quad_format): Likewise.
1352         (c4x_single_format): Likewise.
1353         (c4x_extended_format): Likewise.
1354         (real_internal_format): Likewise.
1355         (coldfire_single_format): New real_format.
1356         (coldfire_double_format): Likewise.
1357         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1358         canonical_nan_lsbs_set.
1359         (pdp11_d_format): Likewise.
1360         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1361         if TARGET_COLDFIRE_CPU.
1362
1363 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1364
1365         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1366         (output_call, m68k_legitimize_call_address): Declare.
1367         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1368         parenthesees.  Add support for a 'W' constraint.
1369         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1370         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1371         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1372         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1373         variables.
1374         (override_options): Initialize them.  Do not set flag_no_function_cse
1375         for TARGET_ID_SHARED_LIBRARY.
1376         (m68k_output_pic_call): Delete.
1377         (m68k_legitimize_call_address): New function.
1378         (print_operand): Remove the %o prefix.  Handle the %p prefix.
1379         (output_call): New function.
1380         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
1381         target address from the GOT if symbolic jumps are not allowed.
1382         * config/m68k/m68k.md (call, general_operand): Do not set
1383         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
1384         Merge separate flag_pic and !flag_pic define_insns into...
1385         (*call, *call_value): ...these new patterns.  Match the address
1386         rather than the containing MEM and require it to be a call_operand.
1387         Use output_call to generate the asm template.
1388         * config/m68k/predicates.md (const_call_operand): New predicate.
1389         (call_operand): Likewise.
1390
1391 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1392
1393         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1394         (STACK_GROWS_DOWNWARD): Define to 1.
1395         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1396         comments, emphasizing that these values are only defaults.
1397         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1398         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1399
1400 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1401
1402         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1403
1404 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1405             Richard Sandiford  <richard@codesourcery.com>
1406
1407         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1408         (FP_REGNO_P): New macros.
1409         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1410         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1411         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1412         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1413
1414 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1415
1416         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1417         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1418         to $tmake_file.
1419         * config/m68k/t-floatlib: New file, extracting common code from...
1420         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1421         * config/m68k/t-uclinux: Here.
1422         * config/m68k/fpgnulib.c: Do not compile extendeed precision
1423         routines on ColdFire targets.
1424
1425 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1426
1427         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1428         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1429         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1430         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1431         explicitly set MOTOROLA to 1.
1432         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1433
1434 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1435             Nathan Sidwell  <nathan@codesourcery.com>
1436
1437         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1438         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1439         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1440         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1441         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1442         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1443         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1444         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1445         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1446         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1447         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1448         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1449         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1450         (DBX_REGISTER_NUMBER): Delete.
1451         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1452         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1453         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1454
1455 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1456
1457         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1458         * config/flat.h: New file.
1459         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1460         OBJECT_FORMAT_FLAT.
1461         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1462         indirect references for -msep-data or -mid-shared-library.
1463         Do not use PC-relative code addresses either.
1464         * config/m68k/m68k.c (override_options): Restrict -fPIC error
1465         to -mpcrel.
1466         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
1467         for shared libraries and crt1.o for executables.  Use crti.o and
1468         crtbegin.o.
1469         (ENDFILE_SPEC): Use crtend.o and crtn.o.
1470         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1471         Do not add -elf2flt or -shared-lib-id options here.
1472         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
1473         Pass -shared-lib-id if -mid-shared-library, taking the library
1474         identifier from -mshared-library-id if given, otherwise
1475         defaulting to 0.
1476         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1477         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1478         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1479         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1480         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1481         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1482         and crtend.o.
1483         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1484         jump sequence for ISA A and ISA A+.
1485         (PICJUMP): Likewise.
1486
1487 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1488
1489         * config.gcc (m68k-*-uclinux*): Base the port on the common
1490         and m68k GNU/Linux files rather than on the generic ELF ones.
1491         * config/m68k/uclinux.h (TARGET_VERSION): Override.
1492         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1493
1494 2007-01-18  Julian Brown  <julian@codesourcery.com>
1495             Richard Sandiford  <richard@codesourcery.com>
1496
1497         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1498         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1499         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1500         before redefining.
1501         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1502         80 unconditionally.
1503         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1504
1505 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1506
1507         * doc/install.texi: Document m68k-uclinuxoldabi.
1508         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1509         * config/m68k/uclinux-oldabi.h: New file, copied from
1510         config/m68k/uclinux.h.
1511
1512 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
1513
1514         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1515         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1516         (m32r_compute_frame_size): Use unsigned for regno.
1517         (m32r_reload_lr): Use gen_frame_mem.
1518         (pop): New.
1519         (m32r_output_function_epilogue): Don't output the function epilogue
1520         textually here.
1521         (m32r_expand_epilogue): New.
1522         (direct_return): Return false if the function has the interrupt
1523         attribute.
1524         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1525         * config/m32r/m32r.md (epilogue): New expander.
1526         (return_lr, return_rte): New insns.
1527         (return): Make it expander.
1528         (return_normal): New expander.
1529
1530 2007-01-18  Josh Conner  <jconner@apple.com>
1531
1532         PR target/30485
1533         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1534         support for UNLE, UNLT, UNGE, and UNGT.
1535
1536 2007-01-18  Jan Hubicka  <jh@suse.cz>
1537
1538         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1539         * tree-tailcall (add_virtual_phis): Likewise.
1540         (optimize_tail_call): Return TODOs.
1541         (execute_tail_calls): Return TODOs.
1542         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1543         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1544         changed.
1545         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1546         Cleanup cfg using TODO.
1547         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1548         * passes.c (execute_function_todo): When cleanup did something, remove
1549         unused locals.
1550         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1551         (make_edges): Don't cleanup_cfg.
1552
1553 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
1554
1555         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1556         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1557         operate on the top of stack.
1558         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1559         register handling of instructions that output to the second
1560         stack slot.
1561         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1562         (move_for_stack_reg): Special-case check for dead destination
1563         stack slot for constant load of 1.0 inside UNSPEC_TAN.
1564
1565         * config/i386/i386.md (UNSPEC_TAN): New constant.
1566         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1567         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1568         to correctly model move of constant 1.0 to top stack slot.
1569         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1570         (unnamed peephole2 pattern): Remove corresponding peephole2
1571         pattern that optimizes tan insn and loading of constant 1.0.
1572         (tanxf2): Use fptanxf4_i387.
1573         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1574         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
1575         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1576         disable patterns for SSE math.
1577         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1578         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1579
1580 2007-01-18  Dirk Mueller  <dmueller@suse.de>
1581             Richard Guenther <rguenther@suse.de>
1582
1583         PR diagnostic/8268
1584         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1585         * common.opt (Warray-bounds): Add new warning option.
1586         * c-opts.c (c_common_handle_option): Define -Warray-bounds
1587         if -Wall is given.
1588         * Makefile.in: make tree-vrp.o depend on toplev.h
1589         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1590         is enabled.
1591         (check_array_refs, check_array_bounds, check_array_ref): New.
1592
1593 2007-01-18  Jan Hubicka  <jh@suse.cz>
1594
1595         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1596         (execute_ssa_ccp): Return flags conditionally.
1597         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1598         changed.
1599         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1600
1601 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
1602
1603         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1604         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1605
1606         * Makefile.c (GTFILES): Add cfglayout.h.
1607         * gengtype.c (open_base_files): Likewise.
1608         * cfglayout.c (cfg_layout_function_footer,
1609         cfg_layout_function_header) Reindent to make gengtype happy.
1610         * cfglayout.h (cfg_layout_function_footer,
1611         cfg_layout_function_header): Add GTY(()) marker.
1612
1613         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1614
1615 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1616
1617         * genautomata.c (write_automata): Include xstrerror output in the
1618         error message if writing the DFA description file fails.
1619
1620 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1621
1622         * config/mips/mips-protos.h (mips_output_external): Make it
1623         return void.
1624         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1625         * config/mips/mips.c (irix_output_external_libcall): Likewise.
1626         (extern_list): Likewise.
1627         (extern_head): Likewise.
1628         (TARGET_ASM_FILE_END): Likewise.
1629         (mips_file_end): Likewise.
1630         (mips_output_external): Rewritten.
1631
1632 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1633
1634         * genpreds.c (write_insn_preds_c): Only write out the function
1635         body for regclass_for_constraint if we have register constraints.
1636
1637 2007-01-17  Tom Tromey  <tromey@redhat.com>
1638
1639         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1640         * doc/install.texi (Testing): Use sourceware.org.
1641         (Binaries): Likewise.
1642         (Specific): Likewise.
1643         * doc/contrib.texi (Contributors): Use sourceware.org.
1644
1645 2007-01-17  Anatoly Sokolov <aesok@post.ru>
1646
1647         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1648         (AVR_ENHANCED): Rename to ...
1649         (AVR_HAVE_MUL): ... new.
1650         (avr_enhanced_p): Rename to ...
1651         (avr_have_mul_p): ... new.
1652         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1653         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1654         * config/avr/avr.c (avr_enhanced_p): Rename to ...
1655         (avr_have_mul_p): ... new.
1656         (base_arch_s): Rename 'enhanced' to 'have_mul'.
1657         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1658         'avr_enhanced_p' and 'enhanced'.
1659         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1660         AVR_HAVE_MUL instead of AVR_ENHANCED.
1661         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1662         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1663         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1664         instead of AVR_ENHANCED.
1665         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1666         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1667         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1668
1669 2007-01-17  Ian Lance Taylor  <iant@google.com>
1670
1671         * vec.h (VEC_reserve_exact): Define.
1672         (vec_gc_p_reserve_exact): Declare.
1673         (vec_gc_o_reserve_exact): Declare.
1674         (vec_heap_p_reserve_exact): Declare.
1675         (vec_heap_o_reserve_exact): Declare.
1676         (VEC_OP (T,A,reserve_exact)): New static inline function, three
1677         versions.
1678         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1679         negative parameter.
1680         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1681         (VEC_OP (T,A,copy)) [all versions]: Likewise.
1682         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1683         * vec.c (calculate_allocation): Add exact parameter.  Change all
1684         callers.
1685         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1686         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1687         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1688         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1689         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1690         (vec_heap_p_reserve_exact): New function.
1691         (vec_heap_o_reserve_exact): New function.
1692
1693 2007-01-17  Jan Hubicka  <jh@suse.cz>
1694
1695         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1696
1697 2007-01-17  Eric Christopher  <echristo@apple.com>
1698
1699         * config.gcc: Support core2 processor.
1700
1701 2007-01-16  Jan Hubicka  <jh@suse.cz>
1702
1703         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1704         the names, just unlink the chain so we don't crash on dangling pointers
1705         to dead SSA names.
1706
1707 2007-01-16  Jan Hubicka  <jh@suse.cz>
1708
1709         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1710         * tree-pass.h: Reorder to make IPA passes appear toegher.
1711         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1712         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1713         parameters, do not call early inliner.
1714         * ipa-inline.c: Update comments.  Include tree-flow.h
1715         (cgraph_decide_inlining): Do not compute inlining parameters.
1716         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1717         be called with function context set up.
1718         (pass_ipa_inline): Remove unreachable functions before pass.
1719         (cgraph_early_inlining): Simplify assuming to be called from the
1720         PM as local pass.
1721         (pass_early_inline): New pass.
1722         (cgraph_gate_ipa_early_inlining): New gate.
1723         (pass_ipa_early_inline): Turn into simple wrapper.
1724         (compute_inline_parameters): New function.
1725         (gate_inline_passes): New gate.
1726         (pass_inline_parameters): New pass.
1727         (apply_inline): Move here from tree-optimize.c
1728         (pass_apply_inline): New pass.
1729         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1730         transforming.
1731         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1732         doing them by hand.
1733         (tree_function_versioning): Do not allocate dummy struct function.
1734         * tree-inline.h (optimize_inline_calls): Update prototype.
1735         * tree-optimize.c (execute_fixup_cfg): Export.
1736         (pass_fixup_cfg): Remove
1737         (tree_rest_of_compilation): Do not apply inlines.
1738         * tree-flow.h (execute_fixup_cfg): Declare.
1739         * Makefile.in (gt-passes.c): New.
1740         * passes.c: Include gt-passes.h
1741         (init_optimization_passes): New passes.
1742         (nnodes, order): New static vars.
1743         (do_per_function_toporder): New function.
1744         (execute_one_pass): Dump current pass here.
1745         (execute_ipa_pass_list): Don't dump current pass here.
1746
1747 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
1748
1749         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1750         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1751
1752 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
1753
1754         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1755         (strong_alias): Define.
1756         (__gcc_qmul): Provide non-FMA for soft-float.
1757         (__gcc_qdiv): Same.
1758         (__gcc_qneg): New.
1759         (__gcc_qeq): New.
1760         (__gcc_qle): New.
1761         (__gcc_qge): New.
1762         (__gcc_qunord): New.
1763         (__gcc_stoq): New.
1764         (__gcc_dtoq): New.
1765         (__gcc_qtos): New.
1766         (__gcc_qtod): New.
1767         (__gcc_qtoi): New.
1768         (__gcc_qtou): New.
1769         (__gcc_itoq): New.
1770         (__gcc_utoq): New.
1771         (fmsub): New.
1772         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1773         soft-float functions.
1774         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1775         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1776         about long double soft float.
1777
1778 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
1779             Tehila Meyzels  <tehila@il.ibm.com>
1780
1781         * tree-vectorizer.h (is_pattern_stmt_p): New.
1782         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1783         formatting (tabs instead of spaces). Cleanup and clarify setting
1784         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1785         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1786         (vectorizable_type_demotion): Check that types are integral.
1787         (vectorizable_type_promotion): Likewise.
1788         (vectorizable_store): Fix typo.  Eliminate new-line at end of
1789         comments.
1790
1791 2007-01-16  Jan Hubicka  <jh@suse.cz>
1792
1793         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1794         cgraph edges too.
1795
1796 2007-01-15  Eric Christopher  <echristo@apple.com>
1797
1798         * ifcvt.c: Include vec.h, vecprim.h.
1799         (check_cond_move_block): New argument regs.
1800         Reorganize. Add registers used to regs.
1801         (cond_move_process_if_block): Use regs set above as
1802         loop bounds.
1803
1804 2007-01-15  Eric Christopher  <echristo@apple.com>
1805
1806         * config/darwin.h: Update copyright.
1807         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1808         (LINK_COMMAND_SPEC): Add -u.
1809         (LINK_SPEC): Fix umbrella for above.
1810
1811 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
1812
1813         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1814         glibc CVS.
1815
1816 2007-01-15  Tom Tromey  <tromey@redhat.com>
1817
1818         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1819         * doc/install.texi (Testing): Don't mention jacks.
1820         (Configuration): Document --enable-java-maintainer-mode.  Move
1821         --with-java-home to libgcj-specific section.  Document
1822         --with-ecj-jar.
1823         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1824
1825 2007-01-15  Jan Hubicka  <jh@suse.cz>
1826
1827         * tree-ssa-dce.c (DCE_TODOs): New.
1828         (propagate_necessity): Return if something changed.
1829         (eliminate_unnecessary_stmts): Likewise.
1830         (perform_tree_ssa_dce): Return TODO flags when needed.
1831         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1832
1833 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
1834
1835         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1836         (fyl2x_extend<mode>xf3_i387): New insn pattern.
1837         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1838         patterns using X87MODEF12 mode macro.  Extend operand 1
1839         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1840         (log10<mode>2): Ditto.
1841         (log2<mode>2): Ditto.
1842         (log1p<mode>2): Ditto.
1843         (logb<mode>2): Ditto.
1844         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1845         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1846         (*fxtractxf3_i387): Rename from *fxtractxf3.
1847         (fxtract_extend<mode>xf3_i387): New insn pattern.
1848         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1849         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1850         and gen_fyl2xxf3_i387().
1851
1852 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
1853
1854         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1855         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1856         get_loop_exit_edges.  Changed return type to bool.
1857         * cfgloop.h (fix_loop_placement): Declaration removed.
1858
1859 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1860
1861         * param.h (MIN_VECT_LOOP_BOUND): New.
1862         * params.def (MIN_VECT_LOOP_BOUND): New.
1863         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1864         argument - minimum threshold for number of iterations.
1865         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1866         argument to declaration.
1867         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1868         MIN_VECT_LOOP_BOUND.
1869         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1870         slpeel_tree_peel_loop_to_edge with additional argument.
1871         (vect_do_peeling_for_alignment): Likewise.
1872         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1873
1874 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1875
1876         PR target/30413
1877         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1878         operands of size 1.
1879
1880 2007-01-14  Jan Hubicka  <jh@suse.cz>
1881
1882         * tree-dfa.c (remove_referenced_var): New function.
1883         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1884         prune referenced vars list too.
1885         * tree-flow.h (remove_referenced_var): Declare.
1886
1887 2007-01-14  Jan Hubicka  <jh@suse.cz>
1888
1889         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1890         separately.
1891         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1892         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1893         * except.h (verify_eh_throw_table_statements): Kill prototype.
1894         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1895         into call exrepssions.
1896         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1897         eh trhow verifier.
1898         * tree-cfg.c: Include pointer-set.h.
1899         (verify_node_sharing): Work on pointer set.
1900         (verify_eh_throw_stmt_node): New.
1901         (verify_stmts): Use pointers sets, verify throw_stmt.
1902
1903 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1904
1905         * ipa-reference.c (analyze_function): Consider also addresses taken
1906         in phi nodes.
1907
1908 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1909
1910         * c-typeck.c (null_pointer_constant_p): Replace use of
1911         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1912         (build_c_cast): Likewise.
1913
1914 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1915
1916         * tree.h (force_fit_type_double): Remove unused final argument.
1917         * c-common.c (constant_expression_warning): Replace use of
1918         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1919         (convert_and_check): Likewise.
1920         (shorten_compare): Update call to force_fit_type_double.
1921         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1922         * convert.c (convert_to_pointer): Update call to
1923         force_fit_type_double.
1924         * fold-const.c (force_fit_type_double): Remove overflowed_const
1925         argument.
1926         (int_const_binop, fold_convert_const_int_from_int,
1927         fold_convert_const_int_from_real, fold_div_compare,
1928         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1929         fold_abs_const, fold_not_const): Remove the final argument from
1930         calls to force_fit_type_double.
1931
1932 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1933
1934         * configure.ac: Set insn to "nop" for spu-*-* also.
1935         * configure: Regenerate.
1936
1937 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
1938
1939         * builtins.def : Add BUILT_IN_FREE.
1940
1941 2007-01-12  Jan Hubicka  <jh@suse.cz>
1942
1943         PR tree-optimization/30443
1944         * tree-inline.c (tree_function_versioning): Do not optimize when
1945         cloning for inlining.
1946
1947 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
1948
1949         * doc/loop.texi: Document recording of loop exits.
1950         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1951         (update_single_exits_after_duplication,
1952         update_single_exit_for_duplicated_loop,
1953         update_single_exit_for_duplicated_loops): Removed.
1954         (duplicate_loop_to_header_edge): Do not call
1955         update_single_exits_after_duplication and
1956         update_single_exit_for_duplicated_loops.
1957         (loop_version): Do not update single_exit information.
1958         (fix_loop_structure): Use record_loop_exits instead of
1959         mark_single_exit_loops.
1960         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1961         the lists of loop exits.
1962         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1963         split_edge, merge_blocks): Update the lists of loop exits.
1964         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1965         loop_optimizer_init.
1966         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1967         of mark_single_exit_loops.
1968         (loop_optimizer_finalize): Call release_recorded_exits.
1969         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1970         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1971         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1972         update single exit information.
1973         * lambda-code.c (perfect_nestify): Ditto.
1974         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1975         (mark_single_exit_loops): Removed.
1976         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1977         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1978         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1979         functions.
1980         (get_loop_exit_edges, single_exit): Use recorded exit lists.
1981         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1982         (verify_loop_structure): Verify consistency of the exit lists.
1983         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
1984         (set_single_exit): Removed.
1985         * cfgloop.h (struct loop_exit): New function.
1986         (struct loop): single_exit_ field replaced by exits field.
1987         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1988         (struct loops): Added exits hash.
1989         (mark_single_exit_loops, set_single_exit): Declaration removed.
1990         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1991
1992 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1993
1994         * doc/invoke.texi: Avoid use of @headitem.
1995
1996 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1997
1998         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1999
2000 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2001
2002         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2003         list of targets.
2004
2005 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2006             Richard Sandiford  <richard@codesourcery.com>
2007
2008         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2009         * config/m68k/m68k.opt: Resort options.
2010         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2011
2012 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2013             Richard Sandiford  <richard@codesourcery.com>
2014
2015         * doc/invoke.texi: Document the macros that are defined by
2016         m68k's -mtune and -mhard-float options.
2017         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2018         (m68k_cpp_cpu_family): Likewise.
2019         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2020         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
2021         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
2022         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2023         (m68k_cpp_cpu_family): Likewise.
2024
2025 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2026
2027         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2028         macros besides mc68000 as tuning macros.  Use a switch statement
2029         to set them and mcpu32.
2030
2031 2007-01-12  Julian Brown  <julian@codesourcery.com>
2032
2033         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2034         (TARGET_68040_ONLY): Rename to...
2035         (TARGET_68040): ...this.
2036         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2037         * config/m68k/m68k.md: Likewise.
2038
2039 2007-01-12  Julian Brown  <julian@codesourcery.com>
2040             Nathan Sidwell  <nathan@codesourcery.com>
2041             Richard Sandiford  <richard@codesourcery.com>
2042
2043         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2044         the -mcpu= argument associated with the --with-cpu setting.
2045         Define M68K_DEFAULT_TUNE to the default -mtune= option,
2046         if different from the one implied by the -mcpu setting.
2047         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2048         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
2049         * doc/invoke.texi: Mention ColdFire in the introduction to the
2050         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
2051         -mno-div and -mhard-float options.  Make -m68881 a synonym for
2052         -mhard-float.  Document the previously-undocumented -m5206e,
2053         -m528x, -m5307 and -m5407 options.  Tweak the existing option
2054         documentation for consistency.
2055         * doc/install.texi: Mention new --with-cpu arguments.
2056         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2057         default CPU if neither -mcpu nor -march are specified.
2058         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2059         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2060         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
2061         using m68k_tune to decide between families that implement the
2062         same ISA.  Use m68k_tune to set __mcfv4e__.
2063         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2064         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2065         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2066         (FL_ISA_C, FL_ISA_MMU): New macros.
2067         (MASK_COLDFIRE): Delete.
2068         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2069         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2070         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2071         (TARGET_HARD_FLOAT): Do not define here.
2072         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2073         (TUNE_68000): New macro.
2074         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2075         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2076         (TUNE_CFV2): Redefine in terms of m68k_tune.
2077         (uarch_type, target_device, fpu_type): New enums.
2078         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2079         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2080         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2081         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2082         (FL_FOR_isa_c): New macros.
2083         (m68k_isa): New enum.
2084         (m68k_target_selection): New structure.
2085         (all_devices, all_isas, all_microarchs): New tables.
2086         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2087         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2088         (MASK_ALL_CPU_BITS): Delete.
2089         (m68k_find_selection): New function.
2090         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2091         Map the legacy target options to a combination of the new ones.
2092         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2093         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
2094         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2095         settings.
2096         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2097         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2098         properties.
2099         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2100         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2101         * config/m68k/m68k-devices.def: New file.
2102
2103 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2104             Nathan Sidwell  <nathan@codesourcery.com>
2105
2106         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2107         (SUBTARGET_EXTRA_SPECS): New macros.
2108         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2109         use %(asm_cpu_spec) instead.
2110         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2111         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2112         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2113         (EXTRA_SPECS): Rename to...
2114         (SUBTARGET_EXTRA_SPECS): ...this.
2115
2116 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2117             Richard Sandiford  <richard@codesourcery.com>
2118             Julian Brown  <julian@codesourcery.com>
2119
2120         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2121         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2122         to the configuration's default CPU.
2123         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2124         Remove default masks.
2125         (m680[012]0-*-*): Set the default with_cpu to the first part of
2126         the target name.
2127         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2128         (m68k*-*-linux): Extend the --with-cpu handling to...
2129         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
2130         and m68010.  Don't set target_cpu_default2.
2131         * doc/install.texi: Document --with-cpu for m68k.
2132         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2133         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2134         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2135         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2136         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2137         (ASM_SPEC): Remove use of %(asm_cpu_default).
2138         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2139         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2140         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2141         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2142         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2143         (TARGET_DEFAULT): Delete.
2144         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2145         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2146         (CPP_CPU_SPEC): Delete.
2147         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2148         the appropriate string.
2149         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2150         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2151         (ASM_SPEC): Don't use %(asm_default_spec).
2152         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2153         TARGET_DEFAULT and add MASK_68881.
2154         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2155
2156 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2157
2158         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2159         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2160         MASK_68010 alongside MASK_68020.
2161         * doc/invoke.texi: Document -m68010.
2162         * config/m68k/m68k.opt (m68010): New.
2163         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2164         if TUNE_68010.
2165         (TUNE_68010): New macro.
2166         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2167         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2168         (M68K_CPU_m68332): Add MASK_68010.
2169         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2170         fallback definition.
2171         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2172         defines.
2173         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2174         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2175         to all entries that use MASK_68020.
2176         (output_move_simode_const, output_move_himode, output_move_qimode)
2177         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2178         instead of TARGET_68020 to select clr behavior.  Remove comment
2179         about there being no TARGET_68010.
2180         * config/m68k/m68k.md: Likewise throughout.
2181
2182 2007-01-12  Julian Brown  <julian@codesourcery.com>
2183
2184         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2185         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2186         * config/m68k/m68k.md: Likewise.
2187
2188 2007-01-12  Julian Brown  <julian@codesourcery.com>
2189
2190         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2191         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2192
2193 2007-01-12  Julian Brown  <julian@codesourcery.com>
2194
2195         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2196         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2197         * config/m68k/m68k.md: Likewise.
2198
2199 2007-01-12  Julian Brown  <julian@codesourcery.com>
2200             Richard Sandiford  <richard@codesourcery.com>
2201
2202         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2203         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2204         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2205         TARGET_CPU32.
2206         (TARGET_CPU32): Rename to...
2207         (TUNE_CPU32): ...this.
2208         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2209         (TUNE_CFV2): New macros.
2210         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2211         remove conditions that are implied by TARGET_68020.
2212         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2213         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2214         (m68k_output_function_epilogue): Likewise.
2215         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2216         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2217         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2218         non-68000 timings.  Refactor multiplication and division costs.
2219         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2220         TUNE_CPU32 instead of TARGET_CPU32.
2221         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2222         and TUNE_68060 instead of TARGET_68060.
2223         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2224         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2225         TARGET_CPU32.
2226         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2227         && !TARGET_COLDFIRE" to choose between moveq and clr.
2228         Likewise in the unnamed movsf pattern.
2229         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2230         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2231         ashiftrt pattern.
2232
2233 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2234
2235         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2236         of tabbing before backslashes.
2237
2238 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2239
2240         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2241         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2242         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2243
2244 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2245
2246         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2247         find_interesting_uses_cond.
2248         (find_interesting_uses_cond): Use extract_cond_operands.
2249         (rewrite_use_compare): Use extract_cond_operands and
2250         force_gimple_operand_bsi.  Do not call update_stmt.
2251         (determine_use_iv_cost_condition): Use extract_cond_operands.
2252         Return cheaper of using original bound and changing the exit bound.
2253
2254 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2255
2256         PR tree-optimization/29516
2257         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2258         most_expensive_mult_to_index, addr_to_parts,
2259         create_mem_ref, maybe_fold_tmr): Make the type of
2260         fields of TARGET_MEM_REF sizetype.
2261         (move_fixed_address_to_symbol, move_pointer_to_base):
2262         New functions.
2263         * tree.def (TARGET_MEM_REF): Add comment on types of
2264         the operands.
2265
2266 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
2267
2268         * c-common.c (vector_types_convertible_p): Treat opaque types as
2269         always convertible if they have the same size, but not otherwise.
2270
2271 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
2272
2273         * ifcvt.c (struct noce_if_info): Add comments to the fields.
2274         Remove the b_unconditional field.
2275         (noce_try_sign_mask): Do not look at b_unconditional.
2276         (noce_process_if_block): Do not use merge_if_blocks.  Update
2277         the CFG here.  Do not set b_unconditional.
2278         (cond_move_process_if_block): Likewise.
2279         (find_cond_trap): Likewise.
2280         (check_cond_move_block): Require simple jump insns at the end
2281         of the basic block.
2282
2283 2007-01-11  Jan Hubicka  <jh@suse.cz>
2284
2285         PR tree-optimization/1046
2286         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2287         when alias info is not ready.
2288         (pass_tail_recursion): Do not require aliasing.
2289         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2290         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2291         * tree-ssa-copy.c (pass_copy_prop): Likewise.
2292         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2293         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2294         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2295         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2296         before inlining.
2297         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2298         When aliasing is not build, mark statement as volatile.
2299
2300 2007-01-11  Tom Tromey  <tromey@redhat.com>
2301
2302         PR preprocessor/15185, PR preprocessor/20989:
2303         * doc/cppopts.texi <-MT>: Update description of algorithm for
2304         computing default target.
2305         <-M, -MD>: Reword "basename" text.
2306
2307 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2308
2309         * builtins.c (expand_builtin_pow, expand_builtin_powi,
2310         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2311         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2312         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2313         fold_builtin_bswap, real_constp, fold_builtin_pow,
2314         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2315         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2316         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2317         * convert.c (convert_to_pointer): Likewise.
2318         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2319         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2320         const_binop, fold_convert_const_int_from_int,
2321         fold_convert_const_int_from_real,
2322         fold_convert_const_real_from_real, sign_bit_p,
2323         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2324         fold_sign_changed_comparison, fold_unary, fold_comparison,
2325         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2326         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2327         * print-tree.c (print_node_brief, print_node): Likewise.
2328         * stor-layout.c (place_field, layout_type): Likewise.
2329         * tree-chrec.c (keep_cast): Likewise.
2330         * tree.c (build_vector, build_real, build_real_from_int_cst,
2331         build_complex): Likewise.
2332
2333 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2334
2335         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
2336         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2337
2338 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
2339
2340         * configure.ac (strict1_warn): Rename to strict_warn.
2341         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2342         cc_set_by_configure, quoted_cc_set_by_configure,
2343         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2344         all_boot_languages, all_stagestuff): Remove.
2345         (target_list): Remove bootstrap targets.
2346         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2347         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2348         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2349         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2350         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2351         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2352         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2353         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2354         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2355         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2356         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2357         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2358         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2359         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2360         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2361         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2362         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2363         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2364         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2365         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2366         (mostlyclean): Adjust.
2367         (clean, distclean): Don't mention bootstrap stuff.
2368         * configure: Regenerate.
2369         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2370         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2371         treelang/config-lang.in (stagestuff): Remove.
2372         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2373         stageprofile, stagefeedback, stagestuff): Remove mention.
2374
2375 2007-01-11  Nick Clifton  <nickc@redhat.com>
2376
2377         * config/mcore/predicates.md (mcore_general_movesrc_operand):
2378         Accept CONSTs.
2379         (mcore_general_movdst_operand): Do not accept CONST_INTs.
2380         (mcore_arith_K_S_operand): Run the test for the S constraint not
2381         the test for the M constraint.
2382         (mcore_addsub_operand): Do not accept integer values that are
2383         larger than 32 bits.
2384         * config/mcore/mcore.md: Remove unused constraints from split.
2385         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2386         (addsi3): Likewise.
2387         (allocate_stack): Likewise.
2388         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2389         operands to 32 bits.
2390         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2391         INTVAL.
2392         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2393         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2394         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2395         mcore_output_bclri, mcore_output_andn, output_inline_const,
2396         mcore_output_move, mcore_output_movedouble): Likewise.
2397         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2398         (output_inline_const): Likewise.
2399         (output_inline_const): Fix format strings used in sprintf
2400         statements.
2401         * config/mcore/mcore-protos.h: Update prototypes for changed
2402         functions in mcore.c.
2403         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2404         HOST_WIDE_INT and not int.
2405         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2406         CONST_OK_FOR_N): Likewise.
2407         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2408         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2409         an INTVAL.
2410
2411 2007-01-10  Jan Hubicka  <jh@suse.cz>
2412
2413         * tree-vrp.c (remove_range_assertions): Release defs.
2414         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2415         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2416
2417 2007-01-10  Paul Brook  <paul@codesourcery.com>
2418
2419         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2420         * config/arm/arm.md (mulsi3subsi): New insn.
2421
2422 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
2423
2424         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2425         (tree_transform_and_unroll_loop): New.
2426         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2427         Declare.
2428
2429 2007-01-10  Robert Kennedy <jimbob@google.com>
2430
2431         * fold-const.c (fold_comparison): Fold comparisons like (x *
2432         1000 < 0) to (x < 0).
2433
2434 2007-01-10  Ian Lance Taylor  <iant@google.com>
2435
2436         * tree-pretty-print.c (dump_generic_node): Print parentheses when
2437         operands have the same priority.
2438
2439 2007-01-10  Tom Tromey  <tromey@redhat.com>
2440
2441         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2442         (fold_binary): Likewise.
2443         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2444         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2445         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2446         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2447         * langhooks.h (struct lang_hooks): Removed field
2448         'can_use_bit_fields_p'.
2449
2450 2007-01-10  Ralf Cors├ępius <ralf.corsepius@rtems.org>
2451
2452         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2453
2454 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
2455
2456         * function.c (get_last_funcdef_no): New function.
2457         * function.h (get_last_funcdef_no): Declare.
2458         * tree-inline.c (initialize_cfun): Add initialization.
2459         (tree_function_versioning): Cleanup.
2460
2461 2007-01-10  Jan Hubicka  <jh@suse.cz>
2462
2463         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2464         PHIs.
2465
2466 2007-01-10  Sa Liu  <saliu@de.ibm.com>
2467             Ben Elliston  <bje@au.ibm.com>
2468
2469         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2470         for SAVE_NONLOCAL and Pmode for any other save level.
2471         * spu-protos.h (spu_restore_stack_block): Declare.
2472         * spu.md (save_stack_block): Remove.
2473         (restore_stack_block): Call spu_restore_stack_block.
2474         * spu.c (spu_restore_stack_block): New function.
2475         (spu_expand_epilogue): Remove old comment.
2476
2477 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
2478
2479         PR tree-optimization/30322
2480         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2481         (cand_value_at): Return the value as aff_tree.
2482         (may_eliminate_iv): Convert the bound from aff_tree to tree.
2483         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2484         aff_combination_mult): New functions.
2485         (aff_combination_add): Use aff_combination_add_cst.
2486         (aff_combination_convert): Allow conversions to a wider type.
2487         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2488         * tree-affine.h (aff_combination_mult): Declare.
2489
2490 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
2491
2492         * doc/tm.texi: Update documentation to reflect reality of exec
2493         and start file search behaviours. Update copyright year.
2494         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2495         header file directories.
2496
2497 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
2498
2499         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2500         (*cosxf2): Rename to cosxf2_i387.
2501         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
2502         using X87MODEF12 mode macro. Rename patterns to
2503         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
2504         for SSE math.
2505         (*cosdf2, *cossf2): Ditto.
2506         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
2507         to match extended input operands.
2508         (sincos<mode>3): New expander.
2509         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2510         insn patterns and corresponding splitters.
2511
2512 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
2513
2514         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2515         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2516
2517 2007-01-09  Nicolas Pitre  <nico@cam.org>
2518
2519         PR target/30173
2520         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2521
2522 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2523
2524         * target.h (struct gcc_target): New field library_rtti_comdat.
2525         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2526         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2527         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2528         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2529
2530 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2531
2532         * doc/invoke.texi (Optimize Options): Correct description of -O0.
2533
2534 2007-01-08  Richard Guenther  <rguenther@suse.de>
2535
2536         * tree.h (force_fit_type_double): Export.
2537         (force_fit_type): Remove.
2538         * fold-const.c (force_fit_type_double): New function.
2539         (force_fit_type): Remove.
2540         (int_const_binop): Use it.
2541         (fold_convert_const_int_from_int): Likewise.
2542         (fold_convert_const_int_from_real): Likewise.
2543         (fold_div_compare): Likewise.
2544         (fold_sign_changed_comparison): Likewise.
2545         (fold_unary): Likewise.
2546         (fold_negate_const): Likewise.
2547         (fold_abs_const): Likewise.
2548         (fold_not_const): Likewise.
2549         * c-common.c (shorten_compare): Use force_fit_type_double.
2550         * convert.c (convert_to_pointer): Likewise.
2551
2552 2007-01-08  Richard Guenther  <rguenther@suse.de>
2553
2554         * tree.h (build_int_cst_wide_type): Export.
2555         * tree.c (build_int_cst_wide_type): New function.
2556         (build_int_cst_wide): Fix comment.
2557         * builtins.c (fold_builtin_object_size): Use build_int_cst
2558         to build -1 or 0 of the correct type.  Use fit_double_type
2559         to check for overflow.
2560         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2561         to build the mask.
2562         (decode_field_reference): Likewise.
2563         (all_ones_mask_p): Likewise.
2564         (native_interpret_int): Use build_int_cst_wide_type.
2565         (fold_binary): Use build_int_cst_type to build an all-ones
2566         value.
2567         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2568
2569 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
2570
2571         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2572
2573 2007-01-09  Ben Elliston  <bje@au.ibm.com>
2574
2575         * genautomata.c (STATS_OPTION): New option.
2576         (stats_flag): New flag.
2577         (gen_automata_option): Handle it.
2578         (initiate_automaton_gen): Ditto.
2579         (write_automata): Output statistics only if stats_flag is
2580         set. Likewise, output time statistics only if time_flag is set.
2581         * doc/md.texi (Processor pipeline description): Document new flag.
2582
2583 2007-01-08  Richard Guenther  <rguenther@suse.de>
2584
2585         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2586         * tree.c (build_int_cst_type): Likewise.
2587         (size_in_bytes): Don't call force_fit_type on the result.
2588         (int_fits_type_p): Use fit_double_type.
2589         * fold-const.c (fit_double_type): New function.
2590         (force_fit_type): Use it.
2591         * tree.h (fit_double_type): Export.
2592
2593 2007-01-08  Jan Hubicka  <jh@suse.cz>
2594
2595         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2596         * ipa.c (function_and_variable_visibility): Fix return type.
2597
2598 2007-01-08  Richard Guenther  <rguenther@suse.de>
2599
2600         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2601         of offset to build the index.
2602         * tree-pretty-print.c (dump_generic_node): Don't build negated
2603         const just for printing.
2604         * c-pretty-print.c (pp_c_integer_constant): Likewise.
2605         * builtins.c (fold_builtin_int_roundingfn): Check if result
2606         fits the type by using force_fit_type and comparing the result.
2607         * predict.c (predict_loops): Use compare_tree_int for comparison.
2608         * tree.c (build_int_cst): Fall back to integer_type_node for
2609         NULL_TREE type.
2610         (build_int_cst_wide): Assert type is non-null.
2611
2612 2007-01-08  Roberto Costa  <roberto.costa@st.com>
2613
2614         * tree-vrp.c (extract_range_from_cond_expr): New.
2615         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2616         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2617         as expressions.
2618         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2619         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2620         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2621         used as expressions.
2622         * tree-object-size.c (cond_expr_object_size): New.
2623         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2624
2625 2007-01-08  Jan Hubicka  <jh@suse.cz>
2626
2627         * tree-ssa-forwprop.c (forward_propagate_into_cond,
2628         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2629         statement.
2630
2631 2007-01-08  Richard Guenther  <rguenther@suse.de>
2632
2633         PR tree-optimization/23603
2634         * tree-vrp.c (set_value_range_to_truthvalue): New function.
2635         (extract_range_from_binary): Fall back to truthvalue instead of
2636         varying for TRUTH_*_EXPR.
2637         (extract_range_from_comparison): Fall back to truthvalue instead of
2638         varying.
2639         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2640         if all visited PHI values were constant.
2641
2642 2007-01-08  Jan Hubicka  <jh@suse.cz>
2643
2644         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2645         (cgraph_analyze_function): break out from ...
2646         (cgraph_finalize_compilation_unit): ... here.
2647         (cgraph_expand_function): Remove forgoten commented out line.
2648         (cgraph_optimize): Analyze functions.
2649
2650 2007-01-08  Jan Hubicka  <jh@suse.cz>
2651
2652         * tree-pas.h (TODO_remove_function): New flag.
2653         (TODO_update*): Renumber.
2654         (pass_ipa_increase_alignment,
2655         pass_ipa_function_and_variable_visibility): New passes.
2656         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2657         (cgraph_function_and_variable_visibility): Move to ipa.c
2658         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2659         cgraph_increase_alignment.
2660         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2661         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2662         calling tree-inline.
2663         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2664         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2665         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2666         (gate_increase_alignment): New function.
2667         (pass_ipa_increase_alignment): New pass.
2668         * ipa.c: Inline tree-pass.h and timevar.h
2669         (function_and_variable_visibility): Move here from cgraphunit.c
2670         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2671         * passes.c (init_optimization_passes): Add the two new passes.
2672         (execute_todo): Handle cgraph_remove_functions.
2673
2674 2007-01-08  Nick Clifton  <nickc@redhat.com>
2675
2676         * config/frv/predicates.md (reg_or_0_operand): Accept
2677         CONST_DOUBLEs.
2678
2679 2007-01-08  Ralf Cors├ępius <ralf.corsepius@rtems.org>
2680
2681         * config/bfin/rtems.h, config/bfin/t-rtems: New.
2682         * config.gcc: Add bfin*-rtems*.
2683
2684 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2685
2686         * c.opt: Add -flax-vector-conversions.
2687         * c-typeck.c (convert_for_assignment): Pass flag to
2688         vector_types_convertible_p to allow emission of note.
2689         (digest_init): Likewise.
2690         * c-opts.c: Handle -flax-vector-conversions.
2691         * c-common.c (flag_lax_vector_conversions): New.
2692         (vector_types_convertible_p): Unless -flax-vector conversions
2693         has been passed, disallow conversions between vectors with
2694         differing numbers of subparts and/or element types.  If such
2695         a conversion is disallowed, possibly emit a note on the first
2696         occasion only to inform the user of -flax-vector-conversions.
2697         The new last argument specifies this.
2698         * c-common.h (flag_lax_vector_conversions): New.
2699         (vector_types_convertible_p): Add extra argument.
2700         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2701         char_type_node for V*QI type vectors.
2702         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2703         Update to satisfy new typechecking rules.
2704         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2705         C and C++ variants.
2706         * doc/invoke.texi (C Dialect Options): Document
2707         -flax-vector-conversions.
2708
2709 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2710
2711         PR tree-optimization/29877
2712         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2713         a register variable on the RHS to not be replaceable.
2714
2715 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
2716         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2717         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2718         const_simm15, const_pow2, const_npow2): Added.
2719         * config/score/misc.md (insv, extv, extzv, movmemsi,
2720         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2721         * config/score/score.c (score_address_cost, score_select_cc_mode):
2722         Added.
2723         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2724         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2725         Update score_print_operand.
2726         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2727         Adjust register allocate order and update some macro define.
2728         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2729         mdx_block_move_straight, mdx_block_move_loop_head,
2730         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2731         mdx_block_move): Added.
2732         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2733         mdp_limm): Updated and fix some bug and typo.
2734         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2735         (movsf, movdf, doloop_end): Added.
2736
2737 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
2738
2739         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2740         config/arm/thumb2.md: Fix comment typos.
2741         * doc/extend.texi: Fix a typo.
2742
2743 2007-01-07  Eric Christopher  <echristo@apple.com>
2744
2745         * configure.ac: Check for __stack_chk_fail for darwin.
2746         * configure: Regenerate.
2747
2748 2007-01-07  Richard Guenther  <rguenther@suse.de>
2749
2750         * tree-vrp.c (extract_range_from_assert): CSE calls to
2751         compare_values where possible.
2752         (extract_range_from_unary_expr): Likewise.
2753
2754 2007-01-07  Anatoly Sokolov <aesok@post.ru>
2755
2756         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2757         * config/avr/avr.c (call_insn_operand): Delete function.
2758         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2759         instead of register number. Use predicates.
2760         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2761         call_insn_operand): Add.
2762
2763 2007-01-06  Jan Hubicka  <jh@suse.cz>
2764
2765         * tree-pass.h (pass_build_cgraph_edges): Declare.
2766         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2767         (visited_nodes): Remove.
2768         (cgraph_create_edges): Move to cgraphbuild.c; rename to
2769         build_cgrpah_edges; make visited_nodes local.
2770         (cgraph_process_new_functions): DO not call initialize_inline_failed.
2771         (record_references_in_initializer): Move to cgraphbuild.c
2772         (initialize_inline_failed, rebuild_cgraph_edges,
2773         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2774         (verify_cgraph_node): Make visited_nodes local.
2775         (cgraph_analyze_function): Do not call cgraph_create_edges and
2776         initialize_inline_failed.
2777         (cgraph_expand_function): Do not call cgraph_lower_function;
2778         assert that function is already lowered.
2779         * Makefile.in (cgraphbuild.o): New.
2780         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2781         at the end of lowering passes.
2782
2783 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
2784
2785         * ifcvt.c (cond_move_convert_if_block): New function, code
2786         factored out from...
2787         (cond_move_process_if_block): ...here.  Call the new function
2788         on the THEN and ELSE blocks.
2789         (merge_if_block): Do not copy global_live_at_end, merge_blocks
2790         already takes care of this.
2791
2792 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2793
2794         PR c/19978
2795         * tree.h (TREE_OVERFLOW_P): New.
2796         * c-typeck.c (parser_build_unary_op): Warn only if result
2797         overflowed and operands did not.
2798         (parser_build_binary_op): Likewise.
2799         (convert_for_assignment): Remove redundant overflow_warning.
2800         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2801
2802 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2803
2804         * c-typeck.c (store_init_value): Split over two lines to follow
2805         the GNU coding style.
2806
2807 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
2808
2809         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2810         __GXX_EXPERIMENTAL_CXX0X__.
2811         * doc/cpp.texi: Same.
2812
2813 2007-01-05  Richard Guenther  <rguenther@suse.de>
2814
2815         PR middle-end/27826
2816         * tree.c (get_narrower): Do not construct COMPONENT_REFs
2817         with mismatched types.  Instead explicitly build a
2818         conversion NOP_EXPR.
2819
2820 2007-01-05  Ian Lance Taylor  <iant@google.com>
2821
2822         * c-common.c (decl_with_nonnull_addr_p): New function.
2823         (c_common_truthvalue_conversion): Call it.
2824         * c-typeck.c (build_binary_op): Likewise.
2825         * c-common.h (decl_with_nonnull_addr_p): Declare.
2826
2827 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
2828
2829         PR c/30360
2830         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2831         denom against 0.0.
2832
2833 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
2834
2835         * doc/install.texi (Final install): Document the fact that
2836         the GNAT runtime should not be stripped.
2837
2838 2007-01-04  Jan Hubicka  <jh@suse.cz>
2839
2840         * tree-inline.c (fold_marked_statements): Update operand caches
2841         and EH after folding
2842
2843 2007-01-04  Ian Lance Taylor  <iant@google.com>
2844
2845         * c-common.c (check_function_nonnull): Whitespace fix.
2846
2847 2007-01-04  Jan Hubicka  <jh@suse.cz>
2848
2849         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2850         comitted older version of patch.
2851         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2852         function, celanup cfg and collect garbage.
2853
2854 2007-01-04  Mike Stump  <mrs@apple.com>
2855
2856         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2857         (clean): Likewise.
2858
2859 2007-01-04  Eric Christopher  <echristo@apple.com>
2860
2861         * libgcc2.c (__bswapsi2): Use SItype.
2862         (__bswapdi2): Use DItype.
2863         * libgcc2.h: Update for above.
2864
2865 2007-01-04  Paul Brook  <paul@codesourcery.com>
2866
2867         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2868         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2869         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2870         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2871         umulsi3_highpart): Make conditional on !arm_arch6.
2872         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2873         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2874         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2875         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2876         umulsi3_highpart_v6): New insns.
2877
2878 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2879
2880         * fold-const.c (fold_convert): When casting an expression to void,
2881         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2882         have a type.  Instead of attempting to build a NOP_EXPR, return
2883         these "special" trees directly.
2884
2885 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2886
2887         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2888         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2889         rs6000_cost->fp not 0 as adjustment for outer NEG.
2890
2891 2007-01-04  Jan Hubicka  <jh@suse.cz>
2892
2893         * cgraph.c (cgraph_release_function_body): New function.
2894         (cgraph_remove_node): Use it.
2895         * cgraph.h (cgraph_release_function_body): Declare.
2896         * cgraphunit.c (cgraph_expand_function): Use it.
2897         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2898         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2899         is initialized and while compilation of other function is running.
2900         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2901         statement CFG annotations from here to ...
2902         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2903
2904 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2905
2906         * cfgloop.h (enum li_flags): Make the constants powers of two.
2907
2908 2007-01-04  Jan Hubicka  <jh@suse.cz>
2909
2910         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2911         set.
2912         (fold_marked_statements): New function.
2913         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2914         * tree-inline.h (copy_body_data): Add statements_to_fold.
2915
2916 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2917
2918         * config.gcc: Mention libgcc/config.host.
2919         * Makefile.in: Update comments mentioning libgcc.
2920         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2921         (all.cross, start.encap, rest.encap, rest.cross): Update
2922         dependencies for libgcc move.
2923         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2924         (clean-target, clean-target-libgcc): Delete.
2925         (srcdirify, GCC_EXTRA_PARTS): New macros.
2926         (libgcc-support, libgcc.mvars): New rules.
2927         (distclean): Remove mention of mklibgcc.
2928         (install): Don't reference INSTALL_LIBGCC.
2929         (install-common): Don't reference EXTRA_PARTS.
2930         (install-libgcc, install-multilib): Delete rules.
2931         * mklibgcc.in: Delete file.
2932         * doc/configfiles.texi: Don't mention mklibgcc.
2933
2934         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2935         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2936         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2937         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2938         (mklibgcc, ldblspecs): Likewise.
2939
2940         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2941         (posixpre.def): Use $(T).
2942         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2943         (SHLIB_SRC, SHLIB_INSTALL): Delete.
2944         (SHLIB_LINK): Make dummy.
2945         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2946
2947         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2948
2949         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2950         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2951         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2952         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2953
2954         * configure.ac (all_outputs): Remove mklibgcc.
2955         * configure: Regenerated.
2956
2957 2007-01-03  Josh Conner  <jconner@apple.com>
2958
2959         PR middle-end/29683
2960         * calls.c (compute_argument_addresses): Set stack and stack_slot
2961         for partial args, too.
2962         (store_one_arg): Use locate.size.constant for the size when
2963         generating a save_area.
2964
2965 2007-01-03  Robert Kennedy <jimbob@google.com>
2966
2967         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2968         whose definitions are deleted due to basic block merging.
2969
2970 2007-01-03  Paul Brook  <paul@codesourcery.com>
2971
2972         PR target/16634
2973         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2974         functions.
2975         (use_return_insn): Return 0 for Thumb interrupt functions.
2976         (print_multi_reg): Add rfe argument for IRQ returns.
2977         (arm_output_epilogue): Pop interrupt return address directly into PC.
2978         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2979
2980 2007-01-03  Paul Brook  <paul@codesourcery.com>
2981
2982         Merge from sourcerygxx-4_1.
2983         * config/arm/thumb2.md: New file.
2984         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2985         Thumb-2.
2986         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2987         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2988         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2989         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2990         tables.
2991         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2992         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2993         comptibility.
2994         * config/arm/ieee754-sf.S: Ditto.
2995         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2996         (thumb1_base_register_rtx_p): ... to this.
2997         (thumb_index_register_rtx_p): Rename...
2998         (thumb1_index_register_rtx_p): ... to this.
2999         (thumb_output_function_prologue): Rename...
3000         (thumb1_output_function_prologue): ... to this.
3001         (thumb_legitimate_address_p): Rename...
3002         (thumb1_legitimate_address_p): ... to this.
3003         (thumb_rtx_costs): Rename...
3004         (thumb1_rtx_costs): ... to this.
3005         (thumb_compute_save_reg_mask): Rename...
3006         (thumb1_compute_save_reg_mask): ... to this.
3007         (thumb_final_prescan_insn): Rename...
3008         (thumb1_final_prescan_insn): ... to this.
3009         (thumb_expand_epilogue): Rename...
3010         (thumb1_expand_epilogue): ... to this.
3011         (arm_unwind_emit_stm): Rename...
3012         (arm_unwind_emit_sequence): ... to this.
3013         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3014         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3015         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3016         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3017         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3018         thumb2_output_casesi): New functions.
3019         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3020         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3021         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3022         THUMB2_WORK_REGS): Define.
3023         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3024         arm_condexec_mask, arm_condexec_masklen)): New variables.
3025         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3026         (arm_override_options): Check new CPU capabilities.
3027         Set new architecture flag variables.
3028         (arm_isr_value): Handle v7m interrupt functions.
3029         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
3030         Thumb-2.
3031         (const_ok_for_arm): Handle Thumb-2 constants.
3032         (arm_gen_constant): Ditto.  Use movw when available.
3033         (arm_function_ok_for_sibcall): Return false for v7m interrupt
3034         functions.
3035         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3036         (thumb_find_work_register, arm_load_pic_register,
3037         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3038         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3039         print_multi_reg, output_mov_long_double_fpa_from_arm,
3040         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3041         output_mov_double_fpa_from_arm, output_move_double,
3042         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3043         output_return_instruction, arm_output_function_prologue,
3044         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3045         arm_output_mi_thunk, thumb_set_return_address): Ditto.
3046         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
3047         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3048         (arithmetic_instr, shift_op): Use arm_shift_nmem.
3049         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
3050         '!' and 'L'.
3051         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3052         (thumb_expand_prologue): Use thumb_set_frame_pointer.
3053         (arm_file_start): Output directive for unified syntax.
3054         (arm_unwind_emit_set): Handle stack alignment instruction.
3055         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3056         Add v6t2, v7, v7a, v7r and v7m.
3057         (RETLDM): Add Thumb-2 code.
3058         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3059         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3060         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3061         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3062         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3063         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3064         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3065         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3066         ADJUST_INSN_LENGTH): Define.
3067         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3068         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3069         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3070         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3071         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3072         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3073         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3074         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3075         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3076         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3077         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3078         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3079         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3080         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3081         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3082         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3083         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3084         cortex-m3.
3085         * config/arm/arm-tune.md: Regenerate.
3086         * config/arm/arm-protos.h: Update prototypes.
3087         * config/arm/vfp.md: Enable patterns for Thumb-2.
3088         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
3089         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3090         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3091         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3092         * config/arm/libunwind.S: Add Thumb-2 code.
3093         * config/arm/constraints.md: Update include Thumb-2.
3094         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3095         * config/arm/ieee754-df.S: Ditto.
3096         * config/arm/bpabi.S: Ditto.
3097         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3098         * config/arm/predicates.md (low_register_operand,
3099         low_reg_or_int_operand, thumb_16bit_operator): New.
3100         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3101         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3102         * config/arm/t-arm-elf: Add armv7 multilib.
3103         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3104         Include thumb2.md.
3105         (UNSPEC_STACK_ALIGN, ce_count): New.
3106         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3107         arm_zero_extendsidi2, arm_zero_extendqidi2): New
3108         insns/expanders.
3109         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3110         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3111         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3112         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3113         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3114         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3115         insns.
3116         * doc/extend.texi: Document ARMv7-M interrupt functions.
3117         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3118
3119 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
3120
3121         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3122         (struct _Unwind_Context): Rename args_size to flags, remove
3123         signal_frame field, add a new args_size field and version field.
3124         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3125         _Unwind_IsExtendedContext): New inline functions.
3126         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3127         Assume by_value array is only present if _Unwind_IsExtendedContext.
3128         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3129         _Unwind_IsSignalFrame.
3130         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3131         (uw_update_context_1): Use _Unwind_SetSignalFrame.
3132         (uw_init_context_1): Initialize context->flags to
3133         EXTENDED_CONTEXT_BIT.
3134         * config/rs6000/linux-unwind.h (frob_update_context): Use
3135         _Unwind_SetSignalFrame.
3136
3137 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3138
3139         PR middle-end/30353
3140         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3141         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3142         the value.
3143
3144 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3145
3146         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3147         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3148         RET, AT_SP): Remove.
3149         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3150         instead of RET.
3151
3152         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3153         * cfgbuild.c (find_basic_blocks): Likewise.
3154         * cfgrtl.c (rtl_create_basic_block): Likewise.
3155         * function.c (temp_slots_at_level): Likewise.
3156         * reg-stack.c (stack_regs_mentioned): Likewise.
3157         * regclass.c (allocate_reg_info): Likewise.
3158         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3159         set_bb_for_stmt, move_block_to_fn): Likewise.
3160         * tree-complex.c (tree_lower_complex): Likewise.
3161         * vec.h (VEC_safe_grow_cleared): New.
3162
3163         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3164
3165 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
3166
3167         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3168         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3169         to loop_version.
3170         * cfgloopmanip.c (scale_loop_frequencies): Export.
3171         (loopify): Scale the frequencies by prescribed coefficients.
3172         (set_zero_probability): New function.
3173         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3174         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3175         and frequencies according to arguments.
3176         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3177         correctly.
3178         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3179         * modulo-sched.c (sms_schedule): Set probabilities for entering
3180         versioned loop correctly.
3181         * tree-vect-transform.c (vect_transform_loop): Ditto.
3182         * cfgloop.h (loopify, loop_version): Declaration changed.
3183         (scale_loop_frequencies): Declared.
3184
3185 2007-01-02  Jan Hubicka  <jh@suse.cz>
3186
3187         * cgraph.c: Include tree-flow.h
3188         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3189         early_local_passes.
3190         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3191         * tree-pass.h (pass_all_early_optimizations): Declare.
3192         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3193         early_local_passes.
3194         (cgraph_analyze_function): Do early_local_passes.
3195         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3196         Do not add referenced vars.
3197         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3198         SSA form.
3199         (gate_all_early_local_passes): New gate.
3200         (pass_early_local_passes): Use new gate.
3201         (execute_early_local_optimizations): New functions.
3202         (gate_all_early_optimizations): New gate.
3203         (pass_all_early_optimizations): New pass.
3204         (execute_free_datastructures): Free SSA only when initialized.
3205         (gate_init_datastructures): Init only when optimizing.
3206         (tree_lowering_passes): Do early local passes when called late.
3207         * tree-profile.c (do_tree_profiling): Don't profile functions added
3208         late.
3209         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3210         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3211         not active.
3212         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3213         during early_local_passes.
3214         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3215
3216
3217 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3218
3219         * Makefile.in: Update copyright year.
3220
3221 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3222
3223         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3224         in $(lang_checks) and check-consistency targets.
3225
3226 2007-01-02  Jan Hubicka  <jh@suse.cz>
3227
3228         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3229         Do not add referenced vars.
3230         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3231         are not active.
3232         * passes.c (init_optimization_passes): Put mudflap_2 after
3233         free_datastructures.
3234
3235 2007-01-02  Jan Hubicka  <jh@suse.cz>
3236
3237         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3238         Set NOTHROW flag on call statements proved to be nothrow.
3239         Update statement of local calls so new pure/const functions are
3240         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3241         goto receivers.
3242         (tree_rest_of_compilation): Register hooks and initialize bitmap
3243         early. Do not set after_inlining flag.
3244
3245 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3246
3247         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3248         HOST_BITS_PER_LONGLONG
3249
3250 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3251
3252         PR c/19977
3253         * c-typeck.c (store_init_value): Don't emit pedantic overflow
3254         warning for non-static initializers.
3255
3256 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
3257
3258         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3259         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3260         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3261         Always use set_unique_reg_note to add REG_EQUAL notes.
3262
3263 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3264
3265         Revert:
3266         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3267
3268         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3269         * cfgbuild.c (find_basic_blocks): Likewise.
3270         * cfgrtl.c (rtl_create_basic_block): Likewise.
3271         * function.c (temp_slots_at_level): Likewise.
3272         * reg-stack.c (stack_regs_mentioned): Likewise.
3273         * regclass.c (allocate_reg_info): Likewise.
3274         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3275         set_bb_for_stmt, move_block_to_fn): Likewise.
3276         * tree-complex.c (tree_lower_complex): Likewise.
3277         * vec.h (VEC_safe_grow_cleared): New.
3278
3279 2007-01-02  Ian Lance Taylor  <iant@google.com>
3280
3281         * c-common.c (c_common_truthvalue_conversion): When warning about
3282         using an assignment as a truth value, set TREE_NO_WARNING.
3283
3284 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3285
3286         PR middle-end/7651
3287         * c.opt (Wold-style-declaration): New.
3288         * doc/invoke.texi (C-only Warning Options): New.
3289         (Wold-style-declaration): Document it.
3290         (Wextra): Enabled by -Wextra.
3291         * c-opts.c (c_common_post_options): Enabled by -Wextra.
3292         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3293         -Wold-style-declaration.
3294
3295 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3296
3297         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3298         * cfgbuild.c (find_basic_blocks): Likewise.
3299         * cfgrtl.c (rtl_create_basic_block): Likewise.
3300         * function.c (temp_slots_at_level): Likewise.
3301         * reg-stack.c (stack_regs_mentioned): Likewise.
3302         * regclass.c (allocate_reg_info): Likewise.
3303         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3304         set_bb_for_stmt, move_block_to_fn): Likewise.
3305         * tree-complex.c (tree_lower_complex): Likewise.
3306         * vec.h (VEC_safe_grow_cleared): New.
3307
3308 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
3309
3310         * c-common.c (c_common_nodes_and_builtins): Since variants of
3311         void_type_node get built before it is given a name, we need to
3312         give those variants the name, too.
3313         (complete_array_type): We need to work with the canonical main
3314         type of the array, from which we will build the qualified version.
3315         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3316         * print-tree.c (print_node): Display canonical type information
3317         for each type.
3318         * stor-layout.c (layout_type): When we don't know the
3319         alignment of a type for which we're building an array, we end up
3320         guessing wrong, so make the type require structural equality.
3321         * tree.c (make_node_stat): When we build a new type, it is its
3322         own canonical type.
3323         (build_type_attribute_qual_variant): When building an attribute
3324         variant, its canonical type is the non-attribute variant. However,
3325         if the attributes are target-dependent and they differ, we need to
3326         use structural equality checks for this type.
3327         (build_qualified_type): A qualified type is not equivalent to its
3328         unqualified variant; set the canonical type appropriately.
3329         (build_distinct_type_copy): When building a distinct type from
3330         another type, the new type is its own canonical type.
3331         (build_variant_type_copy): When building a new type variant, we
3332         assume that it is equivalent to the original type.
3333         (build_pointer_type_for_mode): When building a pointer type, also
3334         build a canonical type pointer.
3335         (build_reference_type_for_mode): When building a reference type,
3336         also build a canonical type reference.
3337         (build_index_type): When we can't hash an index type (e.g.,
3338         because its maximum value is negative), the index type requires
3339         structural equality tests.
3340         (build_array_type): Build the canonical form of an array type.
3341         (build_function_type): Function types require structural equality,
3342         because they contain default arguments, attributes, etc.
3343         (build_method_type_directly): Ditto for method types.
3344         (build_offset_type): Build the canonical offset type.
3345         (build_complex_type): Build the canonical vector type.
3346         (make_vector_type): Build the canonical vector type.
3347         * tree.h (TYPE_CANONICAL): New.
3348         (TYPE_STRUCTURAL_EQUALITY_P): New.
3349         (SET_TYPE_STRUCTURAL_EQUALITY): New.
3350         (struct tree_type): Added "canonical" field.
3351         * params.h (VERIFY_CANONICAL_TYPES): New.
3352         * doc/c-tree.texi (TYPE_CANONICAL): Document.
3353         (TYPE_STRUCTURAL_EQUALITY_P): Document.
3354         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3355         * doc/invoke.texi (verify-canonical-types): Document --param
3356         parameter for verifying canonical types.
3357
3358 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3359
3360         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3361         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3362         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3363         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3364         tm_file.
3365         * config/rs6000/e500.h: New.
3366         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3367         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3368         Remove.
3369         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3370         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3371         TARGET_E500_DOUBLE): Remove.
3372      &nb