gcc/:
[official-gcc.git] / gcc / ChangeLog
blob57598a01cea9f7c64e18976ce1a32e16ba0f0c08
1 2007-01-31  Richard Henderson  <rth@redhat.com>
2             Ian Lance Taylor  <iant@google.com>
4         * lower-subreg.c: New file.
5         * rtl.def (CONCATN): Define.
6         * passes.c (init_optimization_passes): Add pass_lower_subreg and
7         pass_lower_subreg2.
8         * emit-rtl.c (update_reg_offset): New static function, broken out
9         of gen_rtx_REG_offset.
10         (gen_rtx_REG_offset): Call update_reg_offset.
11         (gen_reg_rtx_offset): New function.
12         * regclass.c: Revert patch of 2006-03-05, restoring
13         reg_scan_update.
14         (clear_reg_info_regno): New function.
15         * dwarf2out.c (concatn_loc_descriptor): New static function.
16         (loc_descriptor): Handle CONCATN.
17         * common.opt (fsplit_wide_types): New option.
18         * opts.c (decode_options): Set flag_split_wide_types when
19         optimizing.
20         * timevar.def (TV_LOWER_SUBREG): Define.
21         * rtl.h (gen_reg_rtx_offset): Declare.
22         (reg_scan_update): Declare.
23         * regs.h (clear_reg_info_regno): Declare.
24         * tree-pass.h (pass_lower_subreg): Declare.
25         (pass_lower_subreg2): Declare.
26         * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
27         (Optimize Options): Add -fsplit-wide-types to -O1 list.  Document
28         -fsplit-wide-types.
29         * doc/rtl.texi (Regs and Memory): Document concat and concatn.
30         * Makefile.in (OBJS-common): Add lower-subreg.o.
31         (lower-subreg.o): New target.
33 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
35         * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
37 2007-01-31  Anatoly Sokolov <aesok@post.ru>
39         PR target/19087
40         * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
42 2007-01-31  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
44         PR middle-end/29335
45         * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
47 2007-01-31  Zdenek Dvorak <dvorakz@suse.cz>
49         * cfgloop.h: Include vec-prim.h.
50         (enum li_flags): Remove LI_ONLY_OLD.
51         (loop_iterator): Changed.
52         (fel_next, fel_init): Iterate over loop tree.
53         (FOR_EACH_LOOP_BREAK): New macro.
54         * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
55         FOR_EACH_LOOP.
56         * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
57         * modulo-sched.c (sms_schedule): Ditto.
58         * tree-vectorizer.c (vectorize_loops): Ditto.
59         * doc/loop.texi: Update information on loop numbering and behavior of
60         FOR_EACH_LOOP wrto new loops.
61         * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
62         add_to_evolution_1): Test nestedness of loops instead of comparing
63         their numbers.
64         * tree-chrec.c (chrec_fold_plus_poly_poly,
65         chrec_fold_multiply_poly_poly, chrec_evaluate,
66         hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
67         reset_evolution_in_loop): Ditto.
68         * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
70 2007-01-31  Dirk Mueller  <dmueller@suse.de>
72         * c-common.c (warn_about_parentheses): Separate warning about
73         un-parenthized sequence of comparison operators from the one
74         which is supposed to warn about x <= y <= z.
76 2007-01-31  Uros Bizjak  <ubizjak@gmail.com>
78         * optabs.h (enum optab_index): Add new OTI_isinf.
79         (isinf_optab): Define corresponding macro.
80         * optabs.c (init_optabs): Initialize isinf_optab.
81         * genopinit.c (optabs): Implement isinf_optab using isinf?f2
82         patterns.
83         * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
84         (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
85         using isinf_optab.
86         (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
87         expand_builtin_interclass_mathfn.
88         * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
89         * config/i386/i386.md (UNSPEC_FXAM): New constant.
90         (fxam<mode>2_i387): New insn pattern.
91         (isinf<mode>2) New expander to implement isinf, isinff and isinfl
92         built-in functions as x87 inline asm.
94 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
96         * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
98 2007-01-30  Eric Christopher  <echristo@apple.com>
100         * config.gcc: Add geode.
102 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
104         * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
105         ipa-inline.c, tree-profile.c, tree-ssa-live.c,
106         tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
107         value-prof.c: Fix comment typos.
109 2007-01-30  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
111         PR c++/24745
112         * doc/invoke.texi (Wpointer-arith): Document warning.
114 2007-01-30  Janis Johnson
116         * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
117         the list of discrepancies from the draft TR.
119 2007-01-30  Dirk Mueller  <dmueller@suse.de>
121         PR c++/30601
122         * doc/invoke.texi (-Wreturn-type): Update description to
123         match new behavior.
125 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
127         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
129 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
131         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
132         to handle optabs that operate on floating point input argument and
133         output to integer output.
134         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
135         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
136         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
137         expand_builtin_interclass_mathfn ().
138         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
139         (ilogbsi2): Remove.
140         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
141         ilogbl built-in functions as x87 intrinsics.
143 2007-01-30  Richard Guenther  <rguenther@suse.de>
145         PR middle-end/27657
146         * dwarf2out.c (reference_to_unused): Query varpool if the
147         variable was output.
149 2007-01-30  Richard Guenther  <rguenther@suse.de>
151         PR middle-end/30313
152         * passes.c (execute_one_pass): Reset in_gimple_form to not
153         confuse non-unit-at-a-time mode.
155 2007-01-29  Roger Sayle  <roger@eyesopen.com>
156             Richard Guenther  <rguenther@suse.de>
158         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
159         avoid undefined behaviour on overflow.  Use force_fit_type_double
160         to construct the constant with the specified TREE_OVERFLOW.
162 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
164         * config/dfp-bit.c: Add parameterized support for fp exceptions.
165         * config/dfp-bit.h: Ditto.
167 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
169         * c-decl.c (pop_scope): Replace warnings with call to
170         warn_for_unused_label.
171         * c-common.h (warn_for_unused_label): Declare.
172         * c-common.c (warn_for_unused_label): Define.
174 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
176         * tree-optimize.c (update_inlined_to_pointers): Delete unused
177         function.
179 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
181         * Makefile.in (USER_H): Remove decfloat.h.
182         * ginclude/decfloat.h: Delete, moving contents to ...
183         * ginclude/float.h: Add support for decimal floating point,
184         guarded by __STDC_WANT_DEC_FP__.
186 2007-01-29  Mike Stump  <mrs@apple.com>
188         * doc/gccint.texi (Top): Rename Loop Representation to Loop
189         Analysis and Representation to resolve case insensitive conflict.
190         * doc/loop.texi (Loop Analysis and Representation): Likewise.
192 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
194         * tree.h (struct tree_memory_tag): Add aliases member.
195         (MTAG_ALIASES): New macro.
196         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
197         (add_may_alias): Remove pointer-set. Update for may_aliases being
198         a bitmap.
199         (mark_aliases_call_clobbered): Update for may_aliases being a
200         bitmap.
201         (compute_tag_properties): Ditto.
202         (create_partition_for): Ditto.
203         (compute_memory_partitions): Ditto.
204         (dump_may_aliases_for): Ditto.
205         (is_aliased_with): Ditto.
206         (add_may_alias_for_new_tag): Ditto.
207         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
208         (compute_is_aliased): New function.
209         (compute_may_aliases): Call compute_is_aliased).
210         (init_alias_info): Initialize alias_bitmap_obstack.
211         (union_alias_set_into): New function.
212         (compute_flow_sensitive_aliasing): Use union_aliases_into.
213         (have_common_aliases_p): Rewrite to take two bitmaps and use
214         intersection.
215         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
216         Update for bitmaps.
217         (finalize_ref_all_pointers): Update for add_may_alias changes.
218         (new_type_alias): Ditto.
219         * tree-flow-inline.h (may_aliases): Return a bitmap.
220         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
221         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
222         may_aliases being a bitmap.
223         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
224         may_aliases now returns a bitmap.
225         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
226         being a bitmap.
227         * tree-ssa-operands.c (add_virtual_operand): Update for
228         may_aliases being a bitmap.
230 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
232         PR tree-optimization/30630
233         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
234         variable as changed.
236 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
238         PR c++/28266
239         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
240         expanded only once even if an error occurs.
242 2007-01-29  Ben Elliston  <bje@au.ibm.com>
244         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
245         (__gcov_average_profiler): Likewise.
246         (__gcov_ior_profiler): Likewise.
247         (__gcov_merge_ior): Likewise.
249 2007-01-28  Jan Hubicka  <jh@suse.cz>
251         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
252         * value-prof.c (stringop_block_profile): Likewise.
254 2007-01-28  Jan Hubicka  <jh@suse.cz>
256         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
257         variant handling histograms; add wrapper.
258         (clear_storage_via_libcall): Export.
259         (emit_block_move_hints): Break out from ...; add histograms.
260         (emit_block_move): ... this one.
261         (clear_storage_hints): Break out from ...; add histograms.
262         (clear_storage): ... this one.
263         (set_storage_via_memset): Handle histogram.
264         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
265         (clear_storage_hints, clear_storage_via_libcall): Declare.
266         (set_storage_via_setmem): Update prototype.
267         * doc/md.texi (movmem, setmem): Document new arguments.
269         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
270         new histograms.
271         (stringop_block_profile): New global function.
272         (tree_stringops_values_to_profile): Profile block size and alignment.
273         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
274         HIST_TYPE_IOR.
275         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
276         (stringop_block_profile): Declare.
277         * builtins.c: Include value-prof.h.
278         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
279         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
280         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
281         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
282         * profile.c (instrument_values): Add new counters.
283         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
284         calls.
285         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
286         (tree_init_edge_profiler): Build new profilers.
287         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
288         (pass_tree_profile): Add dump.
289         (tree_profile_hooks): Update.
290         * Makefile.in (LIBGCOV): Add new constants.
291         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
292         __gcov_ior_profiler): New.
293         * i386.md (movmem/setmem expanders): Add new optional arguments.
295 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
297         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
299 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
301         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
302         (tree_unswitch_single_loop): Pass eni_size_weights to
303         tree_num_loop_insns.
304         * tree-ssa-loop-manip.c: Include tree-inline.h.
305         (can_unroll_loop_p): Pass eni_size_weights to
306         tree_num_loop_insns.
307         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
308         Pass eni_size_weights to estimate_num_insns.
309         * tree.h (init_inline_once): Export.
310         * toplev.c (backend_init): Call init_inline_once.
311         * cgraphunit.c (cgraph_process_new_functions,
312         cgraph_analyze_function): Pass eni_inlining_weights to
313         estimate_num_insns.
314         * ipa-inline.c (compute_inline_parameters): Ditto.
315         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
316         to estimate_num_insns.
317         (try_unroll_loop_completely): Pass eni_size_weights to
318         tree_num_loop_insns.
319         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
320         ot estimate_num_insns.
321         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
322         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
323         * tree-inline.c (eni_inlining_weights, eni_size_weights,
324         eni_time_weights): New variables.
325         (init_inline_once): Initialize them.
326         (struct eni_data): Mew.
327         (estimate_num_insns_1, estimate_num_insns): Use weights.
328         * tree-inline.h (struct eni_weights_d): New.
329         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
330         (estimate_num_insns): Declaration changed.
331         * cfgloop.h (tree_num_loop_insns): Declaration changed.
332         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
333         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
335 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
337         * tree-data-ref.c (conflict_fn): Assert that the number of affine
338         relations in the conflict function is valid.
340 2007-01-27  Ian Lance Taylor  <iant@google.com>
342         * common.opt: Add fstrict-overflow.
343         * opts.c (decode_options): Set flag_strict_overflow if -O2.
344         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
345         (TYPE_OVERFLOW_UNDEFINED): Define.
346         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
347         Replace all uses.
348         * tree.h (TYPE_TRAP_SIGNED): Don't define.
349         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
350         (fold_negate_expr): Likewise.
351         (make_range): Likewise.
352         (extract_muldiv_1): Likewise.
353         (maybe_canonicalize_comparison): Likewise.
354         (fold_comparison): Likewise.
355         (fold_binary): Likewise.
356         (tree_expr_nonnegative_p): Likewise.
357         (tree_expr_nonzero_p): Likewise.
358         * tree-vrp.c (compare_values): Likewise.
359         (extract_range_from_binary_expr): Likewise.
360         (extract_range_from_unary_expr): Likewise.
361         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
362         Likewise.
363         (nowrap_type_p): Likewise.
364         * tree-scalar-evolution.c (simple_iv): Likewise.
365         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
366         (build_range_check): Likewise.
367         (extract_muldiv_1): Likewise.
368         (fold_comparison): Likewise.
369         * tree-vrp.c (vrp_int_const_binop): Likewise.
370         (extract_range_from_unary_expr): Likewise.
371         * convert.c (convert_to_integer): Likewise.
372         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
373         (fold_comparison): Likewise.
374         (fold_binary): Likewise.
375         * optabs.c (optab_for_tree_code): Likewise.
376         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
377         * simplify-rtx.c (simplify_const_relational_operation): Check
378         flag_strict_overflow and flag_trapv.
379         (simplify_const_relational_operation): Likewise.
380         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
381         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
382         -fstrict-overflow.
384 2007-01-27  Roger Sayle  <roger@eyesopen.com>
386         * tree.c (tree_fold_gcd): Delete.
387         * tree.h (tree_fold_gcd): Remove prototype.
388         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
389         test whether one constant integer is a multiple of another.  Instead
390         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
391         * fold-const.c (multiple_of_p):  We've determined both TOP and
392         BOTTOM are integer constants so we can call int_const_binop directly
393         instead of the more generic const_binop.
395 2007-01-27  Roger Sayle  <roger@eyesopen.com>
397         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
398         1*X check that the constant hasn't overflowed, to preserve the
399         TREE_OVERFLOW bit.
400         (round_up): Provide an efficient implementation when rouding-up an
401         INTEGER_CST to a power-of-two.
403 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
405         * doc/sourcebuild.texi: Add comma for clarity.
406         * doc/extend.texi: Fix some typos.
407         * doc/passes.texi: Likewise.
408         * doc/cppinternals.texi: Likewise.
409         * doc/c-tree.texi: Likewise.
410         * doc/tree-ssa.texi: Likewise.
411         * doc/install.texi: Likewise.
413 2007-01-27  Jan Hubicka  <jh@suse.cz>
415         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
416         (pass_sra): Do not require alias information.
417         * passes.c (init_optimization_passes): Add SRA
419 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
421         * tracer.c (rest_of_handle_tracer): We already cleaned
422         up the CFG in tracer() so don't do it here again.
423         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
424         cleanup_cfg here, either.  And don't call renumber_insns.
426         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
427         and BB_HEAD are in the insn stream here.  Instead make sure
428         that BB_INSN is valid on all insns.  Also, do check here that
429         there are no pending branch predictions...
430         (rtl_verify_flow_info): ...instead of doing it here.  Checks
431         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
432         here.
434 2007-01-26  Roger Sayle  <roger@eyesopen.com>
436         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
437         function to simplify/factorize operand order canonicalization.
438         (ix86_fixup_binary_operands): Reorganize using the above function.
439         (ix86_binary_operator_ok): Likewise.
441 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
443         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
444         file scope from optimize_attrs.
445         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
446         up search.
447         (optimize_attrs): Clear insn_code_values after freeing it.
449 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
451         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
452         a parts.base assignment.
454 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
456         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
457         (compute_subscript_distance, initialize_data_dependence_relation,
458         finalize_ddr_dependent, analyze_ziv_subscript,
459         analyze_siv_subscript_cst_affine,
460         compute_overlap_steps_for_affine_univar,
461         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
462         analyze_siv_subscript, analyze_miv_subscript,
463         analyze_overlapping_iterations, subscript_dependence_tester_1,
464         compute_self_dependence, free_dependence_relation): Work
465         with affine_fn instead of chrecs.
466         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
467         common_affine_function, affine_function_base,
468         affine_function_constant_p, affine_fn_op, affine_fn_plus,
469         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
470         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
471         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
472         (all_chrecs_equal_p): Removed.
473         * tree-data-ref.h (affine_fn, conflict_function): New types.
474         (struct subscript): Change type of conflicting_iterations_in_a
475         and conflicting_iterations_in_b.
477 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
479         PR other/30182
480         * config/pa/pa.h (TARGET_HPUX_11): New.
481         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
482         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
484 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
485         Richard Guenther  <rguenther@suse.de>
487         * tree-ssa-structalias.c (solve_graph): Handle case
488         we merged the variable to another.
490 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
492         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
493         functions.
495         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
496         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
498 2007-01-25  DJ Delorie  <dj@redhat.com>
500         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
501         allow changes to modes which don't fit in those registers.
503         * reload1.c (choose_reload_regs): Check for invalid subregs before
504         computing their locations, not after.
506 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
508         PR 25127
509         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
510         save Altivec registers in an eh_return function.
511         (compute_vrsave_mask): Likewise.
512         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
513         for saving Altivec registers.
514         (rs6000_emit_prologue): Don't allocate stack twice in
515         eh_return function.  Correct expected value of altivec_save_offset
516         when using save_world.  Describe save of R0 to stack when using
517         save_world.  Describe stack pointer adjustment when using
518         save_world.  Remove duplicated eh_return parameter register saving.
519         Update sp_offset variable after save_world.
520         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
521         darwin-world.asm.
522         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
523         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
524         implies Altivec.
526 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
528         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
530 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
532         PR other/30182
533         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
534         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
536 2007-01-25  Richard Guenther  <rguenther@suse.de>
538         * doc/invoke.texi (-Wcoverage-mismatch): Document.
539         * common.opt (-Wcoverage-mismatch): New warning option.
540         * coverage.c (get_coverage_counts): Ignore coverage mismatch
541         if -Wcoverage-mismatch is given.
543 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
545         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
546         (ipcp_driver): Change to static definition.
547         Add dumping of the ifunctions.
548         (constant_val_insert): Remove unused parameter. Support for SSA.
549         (ipcp_propagate_const): Support for SSA.
550         (ipcp_profile_bb_print): Print only analyzed nodes.
551         (ipcp_replace_map_create): Remove support for Fortran constant
552         for now.
553         * ipa-prop.c (ipa_method_modify_stmt,
554         ipa_callsite_compute_param): Support for SSA.
555         * ipa-prop.h (ipcp_driver): Remove declaration.
556         (IS_VALID_TREE_MAP_INDEX): Add define.
558 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
560         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
562 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
564         PR middle-end/30447
565         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
566         constant argument to cabs and do it without checking for
567         -funsafe-math-optimizations.
569 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
571         * c-common.h (RID_FIRST_CXX0X): New.
572         (RID_LAST_CXX0X): New.
573         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
574         by -Wall.
575         * c.opt (Wc++0x-compat): New.
576         * doc/invoke.texi (-Wc++0x-compat): Document.
578 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
580         * builtins.c (fold_builtin_carg): New.
581         (fold_builtin_1): Use it.
583 2007-01-24  Jan Hubicka  <jh@suse.cz>
585         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
587         * ipa-inline.c (initial_insns, max_insns): Delete.
588         (compute_max_insns): New function.
589         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
590         of insns as base for code growth.
591         (cgraph_decide_inlining): Make initial_insns local; do not compute
592         max_insns.
593         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
594         * doc/invoke.texi (inline-unit-growth): Update docs.
596 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
598         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
599         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
600         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
601         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
602         for CPUs that have PTA_CX16 set.
604         PR middle-end/27416
605         * gimplify.c (omp_check_private): New function.
606         (gimplify_scan_omp_clauses): Use it for
607         firstprivate/lastprivate/reduction.
609         PR middle-end/30494
610         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
611         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
613         PR middle-end/30421
614         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
615         If lower_lastprivate_clauses emits some statements, append them
616         to dlist rather than body_p and to body_p append an initializer.
617         (lower_omp_for): Adjust caller.
619 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
621         * target.h (globalize_decl_name): New.
622         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
623         * output.h (default_globalize_decl_name): New.
624         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
625         globalize_label.
626         (globalize_decl): Ditto.
627         (default_globalize_decl_name): New.
628         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
629         (ia64_handle_version_id_attribute): New.
630         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
631         (ia64_asm_output_external): Use globalize_decl_name instead
632         of globalize_label.
633         * doc/extend.texi (version_id): New pragma.
634         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
636 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
638         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
639         _uleb128_t and _Unwind_SWord with _sleb128_t.
640         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
641         uw_frame_state_for, uw_update_context_1): Likewise.
642         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
643         * unwind-pe.h (read_uleb128, read_sleb128,
644         read_encoded_value_with_base): Likewise.
645         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
647 2007-01-24  Richard Guenther  <rguenther@suse.de>
649         * builtins.c (expand_builtin_cexpi): Get the fndecl
650         for cexp in the correct way.
652 2007-01-24  Jan Hubicka  <jh@suse.cz>
654         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
656 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
658         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
659         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
660         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
661         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
662         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
663         __SIZEOF_PTRDIFF_T__.
664         * doc/cpp.texi: Documentation for the new builtin macros added.
666 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
668         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
669         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
670         instead of operands[2] to avoid access past the end of array.
672 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
674         * reload1.c (emit_reload_insns): Pass the reload register
675         for a non-spill output reload through forget_old_reloads_1.
677 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
679         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
680         _SOFT_DOUBLE if doubles use software floating-point.
681         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
682         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
683         * config/rs6000/darwin-ldouble.c: Also compile functions for
684         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
685         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
686         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
687         commented-out long double override.
688         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
689         * config/rs6000/eabispe.h: Likewise.
690         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
691         long double for non-SPE.
692         (rs6000_handle_option): Likewise.
693         (invalid_e500_subreg): Disallow more subregs involding DImode,
694         DFmode, TImode or TFmode.
695         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
696         E500 double.
697         (legitimate_lo_sum_address_p): Also check for TFmode for E500
698         double.
699         (rs6000_legitimize_address): Also handle TFmode for E500 double.
700         (rs6000_legitimize_reload_address): Also handle TFmode for E500
701         double.
702         (rs6000_legitimate_address): Also check for TFmode for E500
703         double.
704         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
705         (spe_build_register_parallel): Handle TFmode and TCmode.
706         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
707         double.
708         (function_arg): Handle TFmode and TCmode for E500 double.
709         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
710         in general.
711         (print_operand): Handle TFmode and TImode for %y.
712         (rs6000_generate_compare): Handle TFmode comparisons for E500
713         double.
714         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
715         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
716         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
717         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
718         TFmode for E500 double.
719         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
720         (floatsidf2): Enable for E500 double.
721         (movtf_softfloat): Use rs6000_nonimmediate_operand.
722         (extenddftf2): Change to extenddftf2_fprs.
723         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
724         depending on TARGET_E500_DOUBLE.
725         (extendsftf2): Enable for E500 double.
726         (trunctfdf2): Enable for E500 double.
727         (trunctfsf2): Change to trunctfsf2_fprs.
728         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
729         depending on TARGET_E500_DOUBLE.
730         (floatsitf2): Enable for E500 double.
731         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
732         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
733         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
734         (negtf2): Change to negtf2_internal.
735         (negtf2): New expander.
736         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
737         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
738         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
739         (movdi_internal32): Use rs6000_nonimmediate_operand.
740         (unnamed splitter): Likewise.
741         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
742         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
743         (SPE64TF, DITI): New mode macros.
744         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
745         modes.
746         (frob_tf_ti): New.
747         (frob_<mode>_di_2): New.
748         (frob_tf_di_8_2): New.
749         (frob_di_df): Change to frob_di_<mode>; allow more modes.
750         (frob_ti_tf): New.
751         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
752         more modes.
753         (frob_ti_<mode>_8_2): New.
754         (frob_ti_tf_2): New.
755         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
756         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
757         TFmode.
758         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
759         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
760         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
761         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
762         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
763         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
764         tsttflt_gp): New.
766 2007-01-23  Ian Lance Taylor  <iant@google.com>
768         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
769         insn-*.o first.
770         (OBJS-archive): Reformat, alphabetize.
771         (OBJS): Change out_object_file to OBJS-md.
773 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
775         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
776         JUMP_P and CALL_P predicates where applicable.
777         * config/i386/i386.c: Ditto.
778         * config/i386/i386.md: Ditto.
779         * config/i386/mmx.md: Ditto.
780         * config/i386/predicates.md: Ditto.
782 2007-01-22  Andreas Schwab  <schwab@suse.de>
784         * config/m68k/m68k.h: Fix comment.
786 2007-01-22  Jan Hubicka  <jh@suse.cz>
788         * passes.c (init_optimization_passes): Do not rerun
789         pass_early_warn_uninitialized.
791 2007-01-22  Richard Guenther  <rguenther@suse.de>
793         PR tree-optimization/30038
794         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
795         function.
796         (execute_cse_sincos_1): Likewise.
797         (execute_cse_sincos): Likewise.
798         (gate_cse_sincos): Likewise.
799         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
800         the cexpi() canonicalization of sincos().
801         * tree-pass.h (pass_cse_sincos): Declare.
802         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
804 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
806         PR rtl-optimization/29329
807         * combine.c (replaced_rhs_insn): Rename to i2mod.
808         (replaced_rhs_value): Rename to i2mod_new_rhs.
809         (i2mod_old_rhs): New global variable.
810         (combine_instructions): Adjust for above change.  Save a copy of
811         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
812         note are substituted in the second instruction.
813         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
814         ditch the note if it pertains to the second eliminated register
815         and this register is mentioned in i2mod_old_rhs.
817         Revert:
818         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
820         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
821         the insn to which the note was originally attached.
823 2007-01-21  Jan Hubicka  <jh@suse.cz>
825         * ipa-inline.c (inlining_mode): Comment, move up.
826         (cgraph_decide_inlining_incrementally): Do not perform inlining
827         itself; fix handling of flattening of self recursive functions.
828         (cgraph_find_cycles): Remove.
829         (cgraph_flatten_node): Remove.
830         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
831         (try_inline): New function.
832         (cgraph_early_inlining): Update call of
833         cgraph_decide_inlining_incrementally.  Apply inlining here.
834         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
836 2007-01-21  Dirk Mueller  <dmueller@suse.de>
838         PR bootstrap/30511
839         * tree-vrp.c (check_array_bounds): do not warn
840         about ADDR_EXPR's of ARRAY_REF's which are immediately
841         used in binary expressions.
843 2007-01-21  Ira Rosen  <irar@il.ibm.com>
845         * tree-vectorizer.h (struct _stmt_vec_info): Add new field
846         read_write_dep and macros for its access.
847         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
848         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
849         argument, call vect_check_interleaving for every independent pair of
850         data-refs. Mark loads that access the same memory location as a store
851         in the loop.
852         (vect_check_dependences): Remove.
853         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences
854         call, fix the call to vect_analyze_data_ref_dependence.
855         (vect_analyze_data_ref_access): For statements that access the same
856         data-ref, check that they are not stores; for loads, check that there
857         is no store that access the same location.
859 2007-01-20  Roger Sayle  <roger@eyesopen.com>
860             Joseph Myers  <joseph@codesourcery.com>
862         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
863         (-fdump-rtl-stack): Likewise.
864         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
865         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
866         (-mpt-fixed): Use @minus{} for minus sign.
867         (Using Precompiled Headers): Fix under/overfull hbox.
869 2007-01-20  Jan Hubicka  <jh@suse.cz>
871         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
872         place within annotation.
874 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
876         * pa.c (output_move_double): Change array size of xoperands to 4.
878 2007-01-20  Jan Hubicka  <jh@suse.cz>
880         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
881         as temporary.
883 2007-01-19  Ian Lance Taylor  <iant@google.com>
885         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
886         constant to unsigned HOST_WIDE_INT before negating.
888 2007-01-19  Ian Lance Taylor  <iant@google.com>
890         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
891         to unsigned.
892         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
893         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
894         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
895         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
896         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
897         (realloc_vdef, realloc_vuse): Change second parameter to
898         unsigned.
899         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
900         fields to unsigned.
901         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
902         unsigned. Change x and lim locals to unsigned.
903         (realloc_vdef, realloc_vuse): Change num_elem parameter to
904         unsigned.
905         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
906         (copy_virtual_operands): Change i and n locals to unsigned.
908 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
910         PR middle-end/29335
911         * builtins.c (fold_builtin_1): Handle builtin fdim.
913 2007-01-20  Jan Hubicka  <jh@suse.cz>
915         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
916         * tree-ssa-structalias.c (compute_points_to_sets): Do call
917         init_alias_heapvars.
918         (init_alias_heapvars): Initialize only when not already initialized.
919         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
921 2007-01-19  Roger Sayle  <roger@eyesopen.com>
923         * common.opt (fsigned-zeros): New command line option.
924         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
925         of flag_unsafe_math_optimizations.
926         * opts.c (set_fast_math_flags): The -ffast-math command line option
927         implies -fno-signed-zeros.
928         (fast_math_flags_set_p): Likewise.
930         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
931         the documentation of -ffast-math appropriately.  Wrap long lines.
933 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
935         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
936         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
938 2007-01-19  Tomas Bily  <tbily@suse.cz>
940         * cgraphunit.c (cgraph_finalize_function): Updating of pid
941         * tree-profile.c:
942         (tree_init_ic_make_global_vars): New function
943         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
944         (tree_gen_ic_profiler): New function
945         (tree_gen_ic_func_profiler): New function
946         (tree_profiling): Added calling of tree_gen_ic_func_profiler
947         (tree_profile_hooks): Added hook for indirec/virtual calls
948         * value-prof.c (tree_find_values_to_profile): New case for
949         indirect calls
950         (tree_values_to_profile): Call for determining indirect/virtual
951         counters
952         (tree_indirect_call_to_profile): New function
953         (tree_ic_transform): New function
954         (tree_ic): New function
955         (find_func_by_pid): New function
956         (init_pid_map): New function
957         (tree_value_profile_transformations): Added check for
958         indirect/virtual call transformation
959         * value-prof.h (enum hist_type): New counter type for
960         indirect/virtual calls
961         (profile_hooks): Added new hook for profiling indirect/virtual
962         calls
963         * profile.c (instrument_values): New case for indirect/virtual
964         call added
965         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
966         (GCOV_COUNTER_V_INDIR): New counter type
967         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
968         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
969         call added
970         * cgraph.c: Definition of cgraph_max_pid
971         (cgraph_create_node): Default init of pid attribute
972         * cgraph.h: Declaration of cgraph_max_pid
973         (struct cgraph_node): Added pid attribute
974         * libgcov.c (__gcov_indirect_call_profiler): New function
975         (__gcov_one_value_profiler_body): New function
976         (__gcov_one_value_profiler): Body was moved to
977         __gcov_one_value_profiler_body and calls it
979 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
981         * doc/gty.texi (Options): Document the mark_hook option to GTY.
982         * gengtype.c (write_types_data, write_func_for_structure,
983         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
984         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
985         (walk_type, write_func_for_structure): Generate the mark_hook if
986         needed.
988 2007-01-19  Jan Hubicka  <jh@suse.cz>
990         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
991         'early' argument take inlining mode argument specifying whether to
992         inline for size/speeed or all functions; add support for flattening;
993         improve dumpting.
994         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
996 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
998         PR c++/17947
999         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1000         the name of the declared identifier.
1002 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1004         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1005         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1006         duplicate evaluation.
1008 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1010         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1011         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
1012         to XFMode, use acosxf2 and truncate result to requested mode.
1013         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1014         (asin<mode>2): Similarly, with asin expanders.
1015         (*fscalexf4_i387): Rename from *fscalexf4.
1016         (expNcorexf3): New expander.
1017         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1018         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1019         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
1020         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1021         patterns for SSE math.
1022         (exp10<mode>2): Similarly, with exp10 expanders.
1023         (exp2<mode>2): Similarly, with exp2 expanders.
1024         (expm1<mode>2): Similarly, with expm1 expanders.
1025         (ldexp<mode>3): Similarly, with ldexp expanders.
1026         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1027         Use gen_truncxf<mode>2_i387_noop to truncate result.
1029 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1031         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1032         for ColdFire targets.
1034 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
1035             Richard Sandiford  <richard@codesourcery.com>
1037         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1038         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1039         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1040         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1041         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1042         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1044 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1046         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1047         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
1048         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
1049         (*adddi_dilshr32_cf): New pattern.
1050         (adddi3, subdi3): Remove first alternatives.
1052 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1054         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1055         implemented using move.l, do not use its cc result for floating-point
1056         comparisons.
1058 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1060         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1061         from accepting 's' constraints if flag_pic.
1063 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1065         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1066         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1067         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1068         condition from TARGET_68881 to TARGET_HARD_FLOAT.
1070 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
1072         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1073         ColdFire alternatives.
1074         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1075         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
1076         architectures that support ff1 instruction.
1078 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1079             Julian Brown  <julian@codesourcery.com>
1081         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1082         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1083         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
1084         Add commentary.
1086 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1088         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1089         and return a bool.
1090         (output_move_const_into_data_reg, output_move_simode_const): Delete.
1091         * config/m68k/m68k.c (const_method, const_int_cost): Take a
1092         HOST_WIDE_INT instead of an rtx.
1093         (m68k_rtx_costs): Update call accordingly.
1094         (output_move_const_into_data_reg): Likewise.  Fix formatting.
1095         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1096         Return a bool.
1097         (output_move_simode_const): Update calls after above changes.
1098         Rework to use automatic variables and predicates like MEM_P.
1099         * config/m68k/m68k.md (pushexthisi_const): Update call to
1100         valid_mov3q_const.
1102 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1104         * tree-ssa-alias.c (perform_var_substitution): Fix typo
1105         in dump_flags test.
1107 2007-01-19  Richard Guenther  <rguenther@suse.de>
1109         * builtins.c (expand_builtin_cexpi): Fall back to expanding
1110         via cexp in case sincos is not available.
1112 2007-01-19  Richard Guenther  <rguenther@suse.de>
1114         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1115         * defaults.h (TARGET_HAS_SINCOS): Default to off.
1116         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1117         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1118         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1119         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1120         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1121         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1123 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1125         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1126         New insn patterns.
1127         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1128         (atan2xf3): Directly generate RTL pattern.
1129         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1130         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1131         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1132         disable patterns for SSE math.
1133         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1134         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1135         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1136         disable patterns for SSE math.
1138 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1140         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1142 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1144         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1145         VIEW_CONVERT_EXPR to the same type as it's operand.
1147 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1149         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1150         configured for long double 128.
1152 2007-01-18  Mike Stump  <mrs@apple.com>
1154         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1155         error.
1157 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1159         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1160         type correct.
1162 2007-01-18  Jan Hubicka  <jh@suse.cz>
1164         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1165         greater than NUM_VOP_FREE_BUCKETS.
1167 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1169         * tree-ssa-structalias.c: Update comments.
1170         (ptabitmap_obstack): Removed.
1171         (pta_obstack): New.
1172         (oldpta_obstack): Ditto.
1173         (stats): Add a few members.
1174         (struct variable_info): Remove node, complex, address_taken, and
1175         indirect_target members. Add oldsolution member.
1176         (new_var_info): Do not initialize removed members.
1177         (constraint_expr_type): Remove INCLUDES.
1178         (constraint_graph): Add size, implicit_preds, rep,
1179         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1180         (FIRST_REF_NODE): New macro.
1181         (LAST_REF_NODE): Ditto.
1182         (FIRST_ADDR_NODE): Ditto.
1183         (find): New function.
1184         (unite): Ditto.
1185         (dump_constraint): Do not handle INCLUDES.
1186         (insert_into_complex): Do not insert duplicate constraints.
1187         (condense_varmap_nodes): Renamed and rewritten into ...
1188         (merge_node_constraints): This. Also fix bug in handling of
1189         offseted copy constraints.
1190         (clear_edges_for_node): No longer need to deal with preds at all,
1191         or removing associated preds/succs.
1192         (merge_graph_nodes): Deal with indirect_cycles.
1193         Don't deal with predecessors.
1194         (add_implicit_graph_edge): New function.
1195         (add_pred_graph_edge): Ditto.
1196         (add_graph_edge): Don't deal with predecessors.
1197         (build_constraint_graph): Removed.
1198         (build_pred_graph): New function.
1199         (build_succ_graph): Ditto.
1200         (struct scc_info): Removed in_component. Added roots, dfs, and
1201         node_mapping. Remove visited_index, unification_queue.
1202         (scc_visit): Deal with union-find we do now.
1203         Deal with cycles with REF nodes.
1204         (collapse_nodes): Renamed and rewritten to ...
1205         (unify_nodes): This.
1206         (process_unification_queue): Removed.
1207         (topo_visit): Cleanup
1208         (do_da_constraint): Use find.
1209         (do_sd_constraint): Ditto.
1210         (do_ds_constraint): Ditto.
1211         (do_complex_constraint): Ditto.
1212         (init_scc_info): Update for removed and added members.
1213         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1214         (find_indirect_cycles): This.
1215         (equivalence_class): New variable.
1216         (label_visit): New function.
1217         (perform_variable_substitution): Rewritten.
1218         (free_var_substitution_info): New function.
1219         (find_equivalent_node): Ditto.
1220         (move_complex_constraints): Ditto.
1221         (eliminate_indirect_cycles): Ditto.
1222         (solve_graph): Only propagate changed bits.
1223         Use indirect cycle elimination.
1224         Use find.
1225         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1226         (tree_id_eq): Renamed to ...
1227         (tree_vi_eq): This. Update for member change
1228         (insert_id_for_tree): Renamed and rewritten to ...
1229         (insert_vi_for_tree): This.
1230         (lookup_id_for_tree): Renamed and rewritten to ...
1231         (lookup_vi_for_tree): This.
1232         (get_id_for_tree): Renamed and rewritten to ...
1233         (get_vi_for_tree): Ditto.
1234         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1235         (process_constraint): Don't handle INCLUDES.
1236         Remove special ADDRESSOF case.
1237         (find_func_aliases): Rewrite to use vi functions instead of id
1238         ones.
1239         (create_function_info_for): Ditto.
1240         (create_variable_info_for): Ditto.
1241         (intra_create_variable_infos): Ditto.
1242         (merge_smts_into): Ditto.
1243         (find_what_p_points_to): Ditto.
1244         (init_base_vars): Ditto.
1245         (init_alias_vars): Ditto.
1246         (remove_preds_and_fake_succs): New function.
1247         (dump_sa_points_to_info): Dump new stats.
1248         (dump_solution_for_var): Use find.
1249         (set_used_smts): Fix formatting.
1250         (compute_points_to_sets): Updated for new functions.
1251         (ipa_pta_execute): Ditto.
1253 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1254             Richard Sandiford  <richard@codesourcery.com>
1256         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1257         parallels.
1258         * calls.c (expand_call): If the return value is a PARALLEL,
1259         extract its first member.
1260         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1261         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1262         use FUNCTION_EXTRA_EPILOGUE.
1263         (m68k_function_value): Return a PARALLEL if the return value
1264         is of a pointer type.
1265         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1266         (FUNCTION_EXTRA_EPILOGUE): Remove.
1267         * config/m68k/m68k.md (D0_REG): New constant.
1269 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1271         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1272         output a NOP for empty epilogues.
1274 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1276         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1277         before function.  Extend register save check to include all
1278         registers, not just integer ones.
1280 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1282         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1283         (m68k_use_return_insn): ...this.
1284         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1285         * config/m68k/m68k.c (use_return_insn): Rename to...
1286         (m68k_use_return_insn): ...this.
1287         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1288         of USE_RETURN_INSN.
1290 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1292         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1293         nearest even, fix denormal rounding overflow.
1295 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1297         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1298         f%$smove and f%$move.
1299         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1300         (extendsfdf2_cf): Use fdmove instead of f%&move.
1301         (truncdfsf2_cf): Use fsmove instead of f%$smove.
1302         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1304 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1306         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1307         GPR<-GPR moves.
1309 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1311         * real.h (real_format): Add a canonical_nan_lsbs_set field.
1312         (coldfire_single_format): Declare.
1313         (coldfire_double_format): Likewise.
1314         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1315         of qnan_msb_set to determine the lower bits of a canonical
1316         NaN significand.
1317         (encode_ieee_double): Likewise.
1318         (encode_ieee_quad): Likewise.
1319         (ieee_single_format): Initialize canonical_nan_lsbs_set.
1320         (mips_single_format): Likewise.
1321         (ieee_double_format): Likewise.
1322         (mips_double_format): Likewise.
1323         (ieee_extended_motorola_format): Likewise.
1324         (ieee_extended_intel_96_format): Likewise.
1325         (ieee_extended_intel_128_format): Likewise.
1326         (ieee_extended_intel_96_round_53_format): Likewise.
1327         (ibm_extended_format): Likewise.
1328         (mips_extended_format): Likewise.
1329         (ieee_quad_format): Likewise.
1330         (mips_quad_format): Likewise.
1331         (vax_f_format): Likewise.
1332         (vax_d_format): Likewise.
1333         (vax_g_format): Likewise.
1334         (i370_single_format): Likewise.
1335         (i370_double_format): Likewise.
1336         (decimal_single_format): Likewise.
1337         (decimal_double_format): Likewise.
1338         (decimal_quad_format): Likewise.
1339         (c4x_single_format): Likewise.
1340         (c4x_extended_format): Likewise.
1341         (real_internal_format): Likewise.
1342         (coldfire_single_format): New real_format.
1343         (coldfire_double_format): Likewise.
1344         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1345         canonical_nan_lsbs_set.
1346         (pdp11_d_format): Likewise.
1347         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1348         if TARGET_COLDFIRE_CPU.
1350 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1352         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1353         (output_call, m68k_legitimize_call_address): Declare.
1354         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1355         parenthesees.  Add support for a 'W' constraint.
1356         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1357         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1358         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1359         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1360         variables.
1361         (override_options): Initialize them.  Do not set flag_no_function_cse
1362         for TARGET_ID_SHARED_LIBRARY.
1363         (m68k_output_pic_call): Delete.
1364         (m68k_legitimize_call_address): New function.
1365         (print_operand): Remove the %o prefix.  Handle the %p prefix.
1366         (output_call): New function.
1367         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
1368         target address from the GOT if symbolic jumps are not allowed.
1369         * config/m68k/m68k.md (call, general_operand): Do not set
1370         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
1371         Merge separate flag_pic and !flag_pic define_insns into...
1372         (*call, *call_value): ...these new patterns.  Match the address
1373         rather than the containing MEM and require it to be a call_operand.
1374         Use output_call to generate the asm template.
1375         * config/m68k/predicates.md (const_call_operand): New predicate.
1376         (call_operand): Likewise.
1378 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1380         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1381         (STACK_GROWS_DOWNWARD): Define to 1.
1382         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1383         comments, emphasizing that these values are only defaults.
1384         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1385         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1387 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1389         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1391 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1392             Richard Sandiford  <richard@codesourcery.com>
1394         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1395         (FP_REGNO_P): New macros.
1396         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1397         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1398         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1399         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1401 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1403         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1404         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1405         to $tmake_file.
1406         * config/m68k/t-floatlib: New file, extracting common code from...
1407         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1408         * config/m68k/t-uclinux: Here.
1409         * config/m68k/fpgnulib.c: Do not compile extendeed precision
1410         routines on ColdFire targets.
1412 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1414         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1415         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1416         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1417         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1418         explicitly set MOTOROLA to 1.
1419         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1421 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1422             Nathan Sidwell  <nathan@codesourcery.com>
1424         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1425         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1426         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1427         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1428         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1429         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1430         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1431         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1432         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1433         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1434         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1435         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1436         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1437         (DBX_REGISTER_NUMBER): Delete.
1438         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1439         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1440         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1442 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1444         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1445         * config/flat.h: New file.
1446         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1447         OBJECT_FORMAT_FLAT.
1448         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1449         indirect references for -msep-data or -mid-shared-library.
1450         Do not use PC-relative code addresses either.
1451         * config/m68k/m68k.c (override_options): Restrict -fPIC error
1452         to -mpcrel.
1453         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
1454         for shared libraries and crt1.o for executables.  Use crti.o and
1455         crtbegin.o.
1456         (ENDFILE_SPEC): Use crtend.o and crtn.o.
1457         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1458         Do not add -elf2flt or -shared-lib-id options here.
1459         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
1460         Pass -shared-lib-id if -mid-shared-library, taking the library
1461         identifier from -mshared-library-id if given, otherwise
1462         defaulting to 0.
1463         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1464         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1465         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1466         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1467         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1468         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1469         and crtend.o.
1470         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1471         jump sequence for ISA A and ISA A+.
1472         (PICJUMP): Likewise.
1474 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1476         * config.gcc (m68k-*-uclinux*): Base the port on the common
1477         and m68k GNU/Linux files rather than on the generic ELF ones.
1478         * config/m68k/uclinux.h (TARGET_VERSION): Override.
1479         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1481 2007-01-18  Julian Brown  <julian@codesourcery.com>
1482             Richard Sandiford  <richard@codesourcery.com>
1484         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1485         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1486         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1487         before redefining.
1488         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1489         80 unconditionally.
1490         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1492 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1494         * doc/install.texi: Document m68k-uclinuxoldabi.
1495         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1496         * config/m68k/uclinux-oldabi.h: New file, copied from
1497         config/m68k/uclinux.h.
1499 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
1501         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1502         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1503         (m32r_compute_frame_size): Use unsigned for regno.
1504         (m32r_reload_lr): Use gen_frame_mem.
1505         (pop): New.
1506         (m32r_output_function_epilogue): Don't output the function epilogue
1507         textually here.
1508         (m32r_expand_epilogue): New.
1509         (direct_return): Return false if the function has the interrupt
1510         attribute.
1511         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1512         * config/m32r/m32r.md (epilogue): New expander.
1513         (return_lr, return_rte): New insns.
1514         (return): Make it expander.
1515         (return_normal): New expander.
1517 2007-01-18  Josh Conner  <jconner@apple.com>
1519         PR target/30485
1520         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1521         support for UNLE, UNLT, UNGE, and UNGT.
1523 2007-01-18  Jan Hubicka  <jh@suse.cz>
1525         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1526         * tree-tailcall (add_virtual_phis): Likewise.
1527         (optimize_tail_call): Return TODOs.
1528         (execute_tail_calls): Return TODOs.
1529         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1530         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1531         changed.
1532         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1533         Cleanup cfg using TODO.
1534         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1535         * passes.c (execute_function_todo): When cleanup did something, remove
1536         unused locals.
1537         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1538         (make_edges): Don't cleanup_cfg.
1540 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
1542         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1543         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1544         operate on the top of stack.
1545         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1546         register handling of instructions that output to the second
1547         stack slot.
1548         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1549         (move_for_stack_reg): Special-case check for dead destination
1550         stack slot for constant load of 1.0 inside UNSPEC_TAN.
1552         * config/i386/i386.md (UNSPEC_TAN): New constant.
1553         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1554         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1555         to correctly model move of constant 1.0 to top stack slot.
1556         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1557         (unnamed peephole2 pattern): Remove corresponding peephole2
1558         pattern that optimizes tan insn and loading of constant 1.0.
1559         (tanxf2): Use fptanxf4_i387.
1560         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1561         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
1562         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1563         disable patterns for SSE math.
1564         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1565         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1567 2007-01-18  Dirk Mueller  <dmueller@suse.de>
1568             Richard Guenther <rguenther@suse.de>
1570         PR diagnostic/8268
1571         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1572         * common.opt (Warray-bounds): Add new warning option.
1573         * c-opts.c (c_common_handle_option): Define -Warray-bounds
1574         if -Wall is given.
1575         * Makefile.in: make tree-vrp.o depend on toplev.h
1576         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1577         is enabled.
1578         (check_array_refs, check_array_bounds, check_array_ref): New.
1580 2007-01-18  Jan Hubicka  <jh@suse.cz>
1582         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1583         (execute_ssa_ccp): Return flags conditionally.
1584         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1585         changed.
1586         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1588 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
1590         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1591         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1593         * Makefile.c (GTFILES): Add cfglayout.h.
1594         * gengtype.c (open_base_files): Likewise.
1595         * cfglayout.c (cfg_layout_function_footer,
1596         cfg_layout_function_header) Reindent to make gengtype happy.
1597         * cfglayout.h (cfg_layout_function_footer,
1598         cfg_layout_function_header): Add GTY(()) marker.
1600         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1602 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1604         * genautomata.c (write_automata): Include xstrerror output in the
1605         error message if writing the DFA description file fails.
1607 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1609         * config/mips/mips-protos.h (mips_output_external): Make it
1610         return void.
1611         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1612         * config/mips/mips.c (irix_output_external_libcall): Likewise.
1613         (extern_list): Likewise.
1614         (extern_head): Likewise.
1615         (TARGET_ASM_FILE_END): Likewise.
1616         (mips_file_end): Likewise.
1617         (mips_output_external): Rewritten.
1619 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1621         * genpreds.c (write_insn_preds_c): Only write out the function
1622         body for regclass_for_constraint if we have register constraints.
1624 2007-01-17  Tom Tromey  <tromey@redhat.com>
1626         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1627         * doc/install.texi (Testing): Use sourceware.org.
1628         (Binaries): Likewise.
1629         (Specific): Likewise.
1630         * doc/contrib.texi (Contributors): Use sourceware.org.
1632 2007-01-17  Anatoly Sokolov <aesok@post.ru>
1634         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1635         (AVR_ENHANCED): Rename to ...
1636         (AVR_HAVE_MUL): ... new.
1637         (avr_enhanced_p): Rename to ...
1638         (avr_have_mul_p): ... new.
1639         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1640         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1641         * config/avr/avr.c (avr_enhanced_p): Rename to ...
1642         (avr_have_mul_p): ... new.
1643         (base_arch_s): Rename 'enhanced' to 'have_mul'.
1644         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1645         'avr_enhanced_p' and 'enhanced'.
1646         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1647         AVR_HAVE_MUL instead of AVR_ENHANCED.
1648         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1649         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1650         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1651         instead of AVR_ENHANCED.
1652         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1653         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1654         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1656 2007-01-17  Ian Lance Taylor  <iant@google.com>
1658         * vec.h (VEC_reserve_exact): Define.
1659         (vec_gc_p_reserve_exact): Declare.
1660         (vec_gc_o_reserve_exact): Declare.
1661         (vec_heap_p_reserve_exact): Declare.
1662         (vec_heap_o_reserve_exact): Declare.
1663         (VEC_OP (T,A,reserve_exact)): New static inline function, three
1664         versions.
1665         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1666         negative parameter.
1667         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1668         (VEC_OP (T,A,copy)) [all versions]: Likewise.
1669         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1670         * vec.c (calculate_allocation): Add exact parameter.  Change all
1671         callers.
1672         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1673         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1674         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1675         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1676         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1677         (vec_heap_p_reserve_exact): New function.
1678         (vec_heap_o_reserve_exact): New function.
1680 2007-01-17  Jan Hubicka  <jh@suse.cz>
1682         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1684 2007-01-17  Eric Christopher  <echristo@apple.com>
1686         * config.gcc: Support core2 processor.
1688 2007-01-16  Jan Hubicka  <jh@suse.cz>
1690         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1691         the names, just unlink the chain so we don't crash on dangling pointers
1692         to dead SSA names.
1694 2007-01-16  Jan Hubicka  <jh@suse.cz>
1696         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1697         * tree-pass.h: Reorder to make IPA passes appear toegher.
1698         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1699         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1700         parameters, do not call early inliner.
1701         * ipa-inline.c: Update comments.  Include tree-flow.h
1702         (cgraph_decide_inlining): Do not compute inlining parameters.
1703         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1704         be called with function context set up.
1705         (pass_ipa_inline): Remove unreachable functions before pass.
1706         (cgraph_early_inlining): Simplify assuming to be called from the
1707         PM as local pass.
1708         (pass_early_inline): New pass.
1709         (cgraph_gate_ipa_early_inlining): New gate.
1710         (pass_ipa_early_inline): Turn into simple wrapper.
1711         (compute_inline_parameters): New function.
1712         (gate_inline_passes): New gate.
1713         (pass_inline_parameters): New pass.
1714         (apply_inline): Move here from tree-optimize.c
1715         (pass_apply_inline): New pass.
1716         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1717         transforming.
1718         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1719         doing them by hand.
1720         (tree_function_versioning): Do not allocate dummy struct function.
1721         * tree-inline.h (optimize_inline_calls): Update prototype.
1722         * tree-optimize.c (execute_fixup_cfg): Export.
1723         (pass_fixup_cfg): Remove
1724         (tree_rest_of_compilation): Do not apply inlines.
1725         * tree-flow.h (execute_fixup_cfg): Declare.
1726         * Makefile.in (gt-passes.c): New.
1727         * passes.c: Include gt-passes.h
1728         (init_optimization_passes): New passes.
1729         (nnodes, order): New static vars.
1730         (do_per_function_toporder): New function.
1731         (execute_one_pass): Dump current pass here.
1732         (execute_ipa_pass_list): Don't dump current pass here.
1734 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
1736         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1737         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1739 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
1741         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1742         (strong_alias): Define.
1743         (__gcc_qmul): Provide non-FMA for soft-float.
1744         (__gcc_qdiv): Same.
1745         (__gcc_qneg): New.
1746         (__gcc_qeq): New.
1747         (__gcc_qle): New.
1748         (__gcc_qge): New.
1749         (__gcc_qunord): New.
1750         (__gcc_stoq): New.
1751         (__gcc_dtoq): New.
1752         (__gcc_qtos): New.
1753         (__gcc_qtod): New.
1754         (__gcc_qtoi): New.
1755         (__gcc_qtou): New.
1756         (__gcc_itoq): New.
1757         (__gcc_utoq): New.
1758         (fmsub): New.
1759         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1760         soft-float functions.
1761         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1762         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1763         about long double soft float.
1765 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
1766             Tehila Meyzels  <tehila@il.ibm.com>
1768         * tree-vectorizer.h (is_pattern_stmt_p): New.
1769         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1770         formatting (tabs instead of spaces). Cleanup and clarify setting
1771         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1772         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1773         (vectorizable_type_demotion): Check that types are integral.
1774         (vectorizable_type_promotion): Likewise.
1775         (vectorizable_store): Fix typo.  Eliminate new-line at end of
1776         comments.
1778 2007-01-16  Jan Hubicka  <jh@suse.cz>
1780         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1781         cgraph edges too.
1783 2007-01-15  Eric Christopher  <echristo@apple.com>
1785         * ifcvt.c: Include vec.h, vecprim.h.
1786         (check_cond_move_block): New argument regs.
1787         Reorganize. Add registers used to regs.
1788         (cond_move_process_if_block): Use regs set above as
1789         loop bounds.
1791 2007-01-15  Eric Christopher  <echristo@apple.com>
1793         * config/darwin.h: Update copyright.
1794         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1795         (LINK_COMMAND_SPEC): Add -u.
1796         (LINK_SPEC): Fix umbrella for above.
1798 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
1800         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1801         glibc CVS.
1803 2007-01-15  Tom Tromey  <tromey@redhat.com>
1805         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1806         * doc/install.texi (Testing): Don't mention jacks.
1807         (Configuration): Document --enable-java-maintainer-mode.  Move
1808         --with-java-home to libgcj-specific section.  Document
1809         --with-ecj-jar.
1810         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1812 2007-01-15  Jan Hubicka  <jh@suse.cz>
1814         * tree-ssa-dce.c (DCE_TODOs): New.
1815         (propagate_necessity): Return if something changed.
1816         (eliminate_unnecessary_stmts): Likewise.
1817         (perform_tree_ssa_dce): Return TODO flags when needed.
1818         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1820 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
1822         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1823         (fyl2x_extend<mode>xf3_i387): New insn pattern.
1824         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1825         patterns using X87MODEF12 mode macro.  Extend operand 1
1826         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1827         (log10<mode>2): Ditto.
1828         (log2<mode>2): Ditto.
1829         (log1p<mode>2): Ditto.
1830         (logb<mode>2): Ditto.
1831         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1832         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1833         (*fxtractxf3_i387): Rename from *fxtractxf3.
1834         (fxtract_extend<mode>xf3_i387): New insn pattern.
1835         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1836         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1837         and gen_fyl2xxf3_i387().
1839 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
1841         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1842         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1843         get_loop_exit_edges.  Changed return type to bool.
1844         * cfgloop.h (fix_loop_placement): Declaration removed.
1846 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1848         * param.h (MIN_VECT_LOOP_BOUND): New.
1849         * params.def (MIN_VECT_LOOP_BOUND): New.
1850         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1851         argument - minimum threshold for number of iterations.
1852         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1853         argument to declaration.
1854         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1855         MIN_VECT_LOOP_BOUND.
1856         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1857         slpeel_tree_peel_loop_to_edge with additional argument.
1858         (vect_do_peeling_for_alignment): Likewise.
1859         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1861 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1863         PR target/30413
1864         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1865         operands of size 1.
1867 2007-01-14  Jan Hubicka  <jh@suse.cz>
1869         * tree-dfa.c (remove_referenced_var): New function.
1870         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1871         prune referenced vars list too.
1872         * tree-flow.h (remove_referenced_var): Declare.
1874 2007-01-14  Jan Hubicka  <jh@suse.cz>
1876         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1877         separately.
1878         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1879         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1880         * except.h (verify_eh_throw_table_statements): Kill prototype.
1881         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1882         into call exrepssions.
1883         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1884         eh trhow verifier.
1885         * tree-cfg.c: Include pointer-set.h.
1886         (verify_node_sharing): Work on pointer set.
1887         (verify_eh_throw_stmt_node): New.
1888         (verify_stmts): Use pointers sets, verify throw_stmt.
1890 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1892         * ipa-reference.c (analyze_function): Consider also addresses taken
1893         in phi nodes.
1895 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1897         * c-typeck.c (null_pointer_constant_p): Replace use of
1898         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1899         (build_c_cast): Likewise.
1901 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1903         * tree.h (force_fit_type_double): Remove unused final argument.
1904         * c-common.c (constant_expression_warning): Replace use of
1905         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1906         (convert_and_check): Likewise.
1907         (shorten_compare): Update call to force_fit_type_double.
1908         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1909         * convert.c (convert_to_pointer): Update call to
1910         force_fit_type_double.
1911         * fold-const.c (force_fit_type_double): Remove overflowed_const
1912         argument.
1913         (int_const_binop, fold_convert_const_int_from_int,
1914         fold_convert_const_int_from_real, fold_div_compare,
1915         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1916         fold_abs_const, fold_not_const): Remove the final argument from
1917         calls to force_fit_type_double.
1919 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1921         * configure.ac: Set insn to "nop" for spu-*-* also.
1922         * configure: Regenerate.
1924 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
1926         * builtins.def : Add BUILT_IN_FREE.
1928 2007-01-12  Jan Hubicka  <jh@suse.cz>
1930         PR tree-optimization/30443
1931         * tree-inline.c (tree_function_versioning): Do not optimize when
1932         cloning for inlining.
1934 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
1936         * doc/loop.texi: Document recording of loop exits.
1937         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1938         (update_single_exits_after_duplication,
1939         update_single_exit_for_duplicated_loop,
1940         update_single_exit_for_duplicated_loops): Removed.
1941         (duplicate_loop_to_header_edge): Do not call
1942         update_single_exits_after_duplication and
1943         update_single_exit_for_duplicated_loops.
1944         (loop_version): Do not update single_exit information.
1945         (fix_loop_structure): Use record_loop_exits instead of
1946         mark_single_exit_loops.
1947         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1948         the lists of loop exits.
1949         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1950         split_edge, merge_blocks): Update the lists of loop exits.
1951         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1952         loop_optimizer_init.
1953         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1954         of mark_single_exit_loops.
1955         (loop_optimizer_finalize): Call release_recorded_exits.
1956         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1957         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1958         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1959         update single exit information.
1960         * lambda-code.c (perfect_nestify): Ditto.
1961         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1962         (mark_single_exit_loops): Removed.
1963         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1964         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1965         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1966         functions.
1967         (get_loop_exit_edges, single_exit): Use recorded exit lists.
1968         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1969         (verify_loop_structure): Verify consistency of the exit lists.
1970         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
1971         (set_single_exit): Removed.
1972         * cfgloop.h (struct loop_exit): New function.
1973         (struct loop): single_exit_ field replaced by exits field.
1974         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1975         (struct loops): Added exits hash.
1976         (mark_single_exit_loops, set_single_exit): Declaration removed.
1977         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1979 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1981         * doc/invoke.texi: Avoid use of @headitem.
1983 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1985         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1987 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1989         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1990         list of targets.
1992 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1993             Richard Sandiford  <richard@codesourcery.com>
1995         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1996         * config/m68k/m68k.opt: Resort options.
1997         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1999 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2000             Richard Sandiford  <richard@codesourcery.com>
2002         * doc/invoke.texi: Document the macros that are defined by
2003         m68k's -mtune and -mhard-float options.
2004         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2005         (m68k_cpp_cpu_family): Likewise.
2006         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2007         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
2008         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
2009         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2010         (m68k_cpp_cpu_family): Likewise.
2012 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2014         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2015         macros besides mc68000 as tuning macros.  Use a switch statement
2016         to set them and mcpu32.
2018 2007-01-12  Julian Brown  <julian@codesourcery.com>
2020         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2021         (TARGET_68040_ONLY): Rename to...
2022         (TARGET_68040): ...this.
2023         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2024         * config/m68k/m68k.md: Likewise.
2026 2007-01-12  Julian Brown  <julian@codesourcery.com>
2027             Nathan Sidwell  <nathan@codesourcery.com>
2028             Richard Sandiford  <richard@codesourcery.com>
2030         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2031         the -mcpu= argument associated with the --with-cpu setting.
2032         Define M68K_DEFAULT_TUNE to the default -mtune= option,
2033         if different from the one implied by the -mcpu setting.
2034         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2035         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
2036         * doc/invoke.texi: Mention ColdFire in the introduction to the
2037         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
2038         -mno-div and -mhard-float options.  Make -m68881 a synonym for
2039         -mhard-float.  Document the previously-undocumented -m5206e,
2040         -m528x, -m5307 and -m5407 options.  Tweak the existing option
2041         documentation for consistency.
2042         * doc/install.texi: Mention new --with-cpu arguments.
2043         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2044         default CPU if neither -mcpu nor -march are specified.
2045         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2046         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2047         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
2048         using m68k_tune to decide between families that implement the
2049         same ISA.  Use m68k_tune to set __mcfv4e__.
2050         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2051         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2052         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2053         (FL_ISA_C, FL_ISA_MMU): New macros.
2054         (MASK_COLDFIRE): Delete.
2055         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2056         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2057         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2058         (TARGET_HARD_FLOAT): Do not define here.
2059         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2060         (TUNE_68000): New macro.
2061         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2062         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2063         (TUNE_CFV2): Redefine in terms of m68k_tune.
2064         (uarch_type, target_device, fpu_type): New enums.
2065         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2066         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2067         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2068         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2069         (FL_FOR_isa_c): New macros.
2070         (m68k_isa): New enum.
2071         (m68k_target_selection): New structure.
2072         (all_devices, all_isas, all_microarchs): New tables.
2073         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2074         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2075         (MASK_ALL_CPU_BITS): Delete.
2076         (m68k_find_selection): New function.
2077         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2078         Map the legacy target options to a combination of the new ones.
2079         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2080         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
2081         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2082         settings.
2083         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2084         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2085         properties.
2086         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2087         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2088         * config/m68k/m68k-devices.def: New file.
2090 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2091             Nathan Sidwell  <nathan@codesourcery.com>
2093         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2094         (SUBTARGET_EXTRA_SPECS): New macros.
2095         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2096         use %(asm_cpu_spec) instead.
2097         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2098         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2099         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2100         (EXTRA_SPECS): Rename to...
2101         (SUBTARGET_EXTRA_SPECS): ...this.
2103 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2104             Richard Sandiford  <richard@codesourcery.com>
2105             Julian Brown  <julian@codesourcery.com>
2107         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2108         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2109         to the configuration's default CPU.
2110         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2111         Remove default masks.
2112         (m680[012]0-*-*): Set the default with_cpu to the first part of
2113         the target name.
2114         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2115         (m68k*-*-linux): Extend the --with-cpu handling to...
2116         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
2117         and m68010.  Don't set target_cpu_default2.
2118         * doc/install.texi: Document --with-cpu for m68k.
2119         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2120         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2121         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2122         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2123         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2124         (ASM_SPEC): Remove use of %(asm_cpu_default).
2125         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2126         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2127         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2128         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2129         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2130         (TARGET_DEFAULT): Delete.
2131         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2132         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2133         (CPP_CPU_SPEC): Delete.
2134         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2135         the appropriate string.
2136         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2137         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2138         (ASM_SPEC): Don't use %(asm_default_spec).
2139         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2140         TARGET_DEFAULT and add MASK_68881.
2141         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2143 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2145         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2146         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2147         MASK_68010 alongside MASK_68020.
2148         * doc/invoke.texi: Document -m68010.
2149         * config/m68k/m68k.opt (m68010): New.
2150         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2151         if TUNE_68010.
2152         (TUNE_68010): New macro.
2153         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2154         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2155         (M68K_CPU_m68332): Add MASK_68010.
2156         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2157         fallback definition.
2158         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2159         defines.
2160         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2161         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2162         to all entries that use MASK_68020.
2163         (output_move_simode_const, output_move_himode, output_move_qimode)
2164         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2165         instead of TARGET_68020 to select clr behavior.  Remove comment
2166         about there being no TARGET_68010.
2167         * config/m68k/m68k.md: Likewise throughout.
2169 2007-01-12  Julian Brown  <julian@codesourcery.com>
2171         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2172         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2173         * config/m68k/m68k.md: Likewise.
2175 2007-01-12  Julian Brown  <julian@codesourcery.com>
2177         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2178         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2180 2007-01-12  Julian Brown  <julian@codesourcery.com>
2182         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2183         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2184         * config/m68k/m68k.md: Likewise.
2186 2007-01-12  Julian Brown  <julian@codesourcery.com>
2187             Richard Sandiford  <richard@codesourcery.com>
2189         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2190         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2191         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2192         TARGET_CPU32.
2193         (TARGET_CPU32): Rename to...
2194         (TUNE_CPU32): ...this.
2195         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2196         (TUNE_CFV2): New macros.
2197         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2198         remove conditions that are implied by TARGET_68020.
2199         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2200         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2201         (m68k_output_function_epilogue): Likewise.
2202         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2203         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2204         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2205         non-68000 timings.  Refactor multiplication and division costs.
2206         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2207         TUNE_CPU32 instead of TARGET_CPU32.
2208         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2209         and TUNE_68060 instead of TARGET_68060.
2210         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2211         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2212         TARGET_CPU32.
2213         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2214         && !TARGET_COLDFIRE" to choose between moveq and clr.
2215         Likewise in the unnamed movsf pattern.
2216         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2217         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2218         ashiftrt pattern.
2220 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2222         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2223         of tabbing before backslashes.
2225 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2227         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2228         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2229         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2231 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2233         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2234         find_interesting_uses_cond.
2235         (find_interesting_uses_cond): Use extract_cond_operands.
2236         (rewrite_use_compare): Use extract_cond_operands and
2237         force_gimple_operand_bsi.  Do not call update_stmt.
2238         (determine_use_iv_cost_condition): Use extract_cond_operands.
2239         Return cheaper of using original bound and changing the exit bound.
2241 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2243         PR tree-optimization/29516
2244         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2245         most_expensive_mult_to_index, addr_to_parts,
2246         create_mem_ref, maybe_fold_tmr): Make the type of
2247         fields of TARGET_MEM_REF sizetype.
2248         (move_fixed_address_to_symbol, move_pointer_to_base):
2249         New functions.
2250         * tree.def (TARGET_MEM_REF): Add comment on types of
2251         the operands.
2253 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
2255         * c-common.c (vector_types_convertible_p): Treat opaque types as
2256         always convertible if they have the same size, but not otherwise.
2258 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
2260         * ifcvt.c (struct noce_if_info): Add comments to the fields.
2261         Remove the b_unconditional field.
2262         (noce_try_sign_mask): Do not look at b_unconditional.
2263         (noce_process_if_block): Do not use merge_if_blocks.  Update
2264         the CFG here.  Do not set b_unconditional.
2265         (cond_move_process_if_block): Likewise.
2266         (find_cond_trap): Likewise.
2267         (check_cond_move_block): Require simple jump insns at the end
2268         of the basic block.
2270 2007-01-11  Jan Hubicka  <jh@suse.cz>
2272         PR tree-optimization/1046
2273         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2274         when alias info is not ready.
2275         (pass_tail_recursion): Do not require aliasing.
2276         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2277         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2278         * tree-ssa-copy.c (pass_copy_prop): Likewise.
2279         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2280         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2281         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2282         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2283         before inlining.
2284         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2285         When aliasing is not build, mark statement as volatile.
2287 2007-01-11  Tom Tromey  <tromey@redhat.com>
2289         PR preprocessor/15185, PR preprocessor/20989:
2290         * doc/cppopts.texi <-MT>: Update description of algorithm for
2291         computing default target.
2292         <-M, -MD>: Reword "basename" text.
2294 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2296         * builtins.c (expand_builtin_pow, expand_builtin_powi,
2297         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2298         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2299         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2300         fold_builtin_bswap, real_constp, fold_builtin_pow,
2301         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2302         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2303         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2304         * convert.c (convert_to_pointer): Likewise.
2305         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2306         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2307         const_binop, fold_convert_const_int_from_int,
2308         fold_convert_const_int_from_real,
2309         fold_convert_const_real_from_real, sign_bit_p,
2310         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2311         fold_sign_changed_comparison, fold_unary, fold_comparison,
2312         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2313         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2314         * print-tree.c (print_node_brief, print_node): Likewise.
2315         * stor-layout.c (place_field, layout_type): Likewise.
2316         * tree-chrec.c (keep_cast): Likewise.
2317         * tree.c (build_vector, build_real, build_real_from_int_cst,
2318         build_complex): Likewise.
2320 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2322         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
2323         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2325 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
2327         * configure.ac (strict1_warn): Rename to strict_warn.
2328         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2329         cc_set_by_configure, quoted_cc_set_by_configure,
2330         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2331         all_boot_languages, all_stagestuff): Remove.
2332         (target_list): Remove bootstrap targets.
2333         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2334         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2335         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2336         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2337         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2338         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2339         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2340         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2341         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2342         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2343         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2344         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2345         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2346         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2347         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2348         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2349         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2350         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2351         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2352         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2353         (mostlyclean): Adjust.
2354         (clean, distclean): Don't mention bootstrap stuff.
2355         * configure: Regenerate.
2356         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2357         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2358         treelang/config-lang.in (stagestuff): Remove.
2359         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2360         stageprofile, stagefeedback, stagestuff): Remove mention.
2362 2007-01-11  Nick Clifton  <nickc@redhat.com>
2364         * config/mcore/predicates.md (mcore_general_movesrc_operand):
2365         Accept CONSTs.
2366         (mcore_general_movdst_operand): Do not accept CONST_INTs.
2367         (mcore_arith_K_S_operand): Run the test for the S constraint not
2368         the test for the M constraint.
2369         (mcore_addsub_operand): Do not accept integer values that are
2370         larger than 32 bits.
2371         * config/mcore/mcore.md: Remove unused constraints from split.
2372         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2373         (addsi3): Likewise.
2374         (allocate_stack): Likewise.
2375         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2376         operands to 32 bits.
2377         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2378         INTVAL.
2379         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2380         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2381         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2382         mcore_output_bclri, mcore_output_andn, output_inline_const,
2383         mcore_output_move, mcore_output_movedouble): Likewise.
2384         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2385         (output_inline_const): Likewise.
2386         (output_inline_const): Fix format strings used in sprintf
2387         statements.
2388         * config/mcore/mcore-protos.h: Update prototypes for changed
2389         functions in mcore.c.
2390         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2391         HOST_WIDE_INT and not int.
2392         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2393         CONST_OK_FOR_N): Likewise.
2394         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2395         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2396         an INTVAL.
2398 2007-01-10  Jan Hubicka  <jh@suse.cz>
2400         * tree-vrp.c (remove_range_assertions): Release defs.
2401         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2402         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2404 2007-01-10  Paul Brook  <paul@codesourcery.com>
2406         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2407         * config/arm/arm.md (mulsi3subsi): New insn.
2409 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
2411         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2412         (tree_transform_and_unroll_loop): New.
2413         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2414         Declare.
2416 2007-01-10  Robert Kennedy <jimbob@google.com>
2418         * fold-const.c (fold_comparison): Fold comparisons like (x *
2419         1000 < 0) to (x < 0).
2421 2007-01-10  Ian Lance Taylor  <iant@google.com>
2423         * tree-pretty-print.c (dump_generic_node): Print parentheses when
2424         operands have the same priority.
2426 2007-01-10  Tom Tromey  <tromey@redhat.com>
2428         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2429         (fold_binary): Likewise.
2430         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2431         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2432         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2433         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2434         * langhooks.h (struct lang_hooks): Removed field
2435         'can_use_bit_fields_p'.
2437 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
2439         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2441 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
2443         * function.c (get_last_funcdef_no): New function.
2444         * function.h (get_last_funcdef_no): Declare.
2445         * tree-inline.c (initialize_cfun): Add initialization.
2446         (tree_function_versioning): Cleanup.
2448 2007-01-10  Jan Hubicka  <jh@suse.cz>
2450         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2451         PHIs.
2453 2007-01-10  Sa Liu  <saliu@de.ibm.com>
2454             Ben Elliston  <bje@au.ibm.com>
2456         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2457         for SAVE_NONLOCAL and Pmode for any other save level.
2458         * spu-protos.h (spu_restore_stack_block): Declare.
2459         * spu.md (save_stack_block): Remove.
2460         (restore_stack_block): Call spu_restore_stack_block.
2461         * spu.c (spu_restore_stack_block): New function.
2462         (spu_expand_epilogue): Remove old comment.
2464 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
2466         PR tree-optimization/30322
2467         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2468         (cand_value_at): Return the value as aff_tree.
2469         (may_eliminate_iv): Convert the bound from aff_tree to tree.
2470         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2471         aff_combination_mult): New functions.
2472         (aff_combination_add): Use aff_combination_add_cst.
2473         (aff_combination_convert): Allow conversions to a wider type.
2474         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2475         * tree-affine.h (aff_combination_mult): Declare.
2477 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
2479         * doc/tm.texi: Update documentation to reflect reality of exec
2480         and start file search behaviours. Update copyright year.
2481         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2482         header file directories.
2484 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
2486         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2487         (*cosxf2): Rename to cosxf2_i387.
2488         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
2489         using X87MODEF12 mode macro. Rename patterns to
2490         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
2491         for SSE math.
2492         (*cosdf2, *cossf2): Ditto.
2493         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
2494         to match extended input operands.
2495         (sincos<mode>3): New expander.
2496         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2497         insn patterns and corresponding splitters.
2499 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
2501         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2502         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2504 2007-01-09  Nicolas Pitre  <nico@cam.org>
2506         PR target/30173
2507         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2509 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2511         * target.h (struct gcc_target): New field library_rtti_comdat.
2512         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2513         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2514         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2515         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2517 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2519         * doc/invoke.texi (Optimize Options): Correct description of -O0.
2521 2007-01-08  Richard Guenther  <rguenther@suse.de>
2523         * tree.h (force_fit_type_double): Export.
2524         (force_fit_type): Remove.
2525         * fold-const.c (force_fit_type_double): New function.
2526         (force_fit_type): Remove.
2527         (int_const_binop): Use it.
2528         (fold_convert_const_int_from_int): Likewise.
2529         (fold_convert_const_int_from_real): Likewise.
2530         (fold_div_compare): Likewise.
2531         (fold_sign_changed_comparison): Likewise.
2532         (fold_unary): Likewise.
2533         (fold_negate_const): Likewise.
2534         (fold_abs_const): Likewise.
2535         (fold_not_const): Likewise.
2536         * c-common.c (shorten_compare): Use force_fit_type_double.
2537         * convert.c (convert_to_pointer): Likewise.
2539 2007-01-08  Richard Guenther  <rguenther@suse.de>
2541         * tree.h (build_int_cst_wide_type): Export.
2542         * tree.c (build_int_cst_wide_type): New function.
2543         (build_int_cst_wide): Fix comment.
2544         * builtins.c (fold_builtin_object_size): Use build_int_cst
2545         to build -1 or 0 of the correct type.  Use fit_double_type
2546         to check for overflow.
2547         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2548         to build the mask.
2549         (decode_field_reference): Likewise.
2550         (all_ones_mask_p): Likewise.
2551         (native_interpret_int): Use build_int_cst_wide_type.
2552         (fold_binary): Use build_int_cst_type to build an all-ones
2553         value.
2554         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2556 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
2558         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2560 2007-01-09  Ben Elliston  <bje@au.ibm.com>
2562         * genautomata.c (STATS_OPTION): New option.
2563         (stats_flag): New flag.
2564         (gen_automata_option): Handle it.
2565         (initiate_automaton_gen): Ditto.
2566         (write_automata): Output statistics only if stats_flag is
2567         set. Likewise, output time statistics only if time_flag is set.
2568         * doc/md.texi (Processor pipeline description): Document new flag.
2570 2007-01-08  Richard Guenther  <rguenther@suse.de>
2572         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2573         * tree.c (build_int_cst_type): Likewise.
2574         (size_in_bytes): Don't call force_fit_type on the result.
2575         (int_fits_type_p): Use fit_double_type.
2576         * fold-const.c (fit_double_type): New function.
2577         (force_fit_type): Use it.
2578         * tree.h (fit_double_type): Export.
2580 2007-01-08  Jan Hubicka  <jh@suse.cz>
2582         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2583         * ipa.c (function_and_variable_visibility): Fix return type.
2585 2007-01-08  Richard Guenther  <rguenther@suse.de>
2587         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2588         of offset to build the index.
2589         * tree-pretty-print.c (dump_generic_node): Don't build negated
2590         const just for printing.
2591         * c-pretty-print.c (pp_c_integer_constant): Likewise.
2592         * builtins.c (fold_builtin_int_roundingfn): Check if result
2593         fits the type by using force_fit_type and comparing the result.
2594         * predict.c (predict_loops): Use compare_tree_int for comparison.
2595         * tree.c (build_int_cst): Fall back to integer_type_node for
2596         NULL_TREE type.
2597         (build_int_cst_wide): Assert type is non-null.
2599 2007-01-08  Roberto Costa  <roberto.costa@st.com>
2601         * tree-vrp.c (extract_range_from_cond_expr): New.
2602         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2603         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2604         as expressions.
2605         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2606         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2607         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2608         used as expressions.
2609         * tree-object-size.c (cond_expr_object_size): New.
2610         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2612 2007-01-08  Jan Hubicka  <jh@suse.cz>
2614         * tree-ssa-forwprop.c (forward_propagate_into_cond,
2615         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2616         statement.
2618 2007-01-08  Richard Guenther  <rguenther@suse.de>
2620         PR tree-optimization/23603
2621         * tree-vrp.c (set_value_range_to_truthvalue): New function.
2622         (extract_range_from_binary): Fall back to truthvalue instead of
2623         varying for TRUTH_*_EXPR.
2624         (extract_range_from_comparison): Fall back to truthvalue instead of
2625         varying.
2626         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2627         if all visited PHI values were constant.
2629 2007-01-08  Jan Hubicka  <jh@suse.cz>
2631         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2632         (cgraph_analyze_function): break out from ...
2633         (cgraph_finalize_compilation_unit): ... here.
2634         (cgraph_expand_function): Remove forgoten commented out line.
2635         (cgraph_optimize): Analyze functions.
2637 2007-01-08  Jan Hubicka  <jh@suse.cz>
2639         * tree-pas.h (TODO_remove_function): New flag.
2640         (TODO_update*): Renumber.
2641         (pass_ipa_increase_alignment,
2642         pass_ipa_function_and_variable_visibility): New passes.
2643         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2644         (cgraph_function_and_variable_visibility): Move to ipa.c
2645         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2646         cgraph_increase_alignment.
2647         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2648         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2649         calling tree-inline.
2650         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2651         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2652         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2653         (gate_increase_alignment): New function.
2654         (pass_ipa_increase_alignment): New pass.
2655         * ipa.c: Inline tree-pass.h and timevar.h
2656         (function_and_variable_visibility): Move here from cgraphunit.c
2657         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2658         * passes.c (init_optimization_passes): Add the two new passes.
2659         (execute_todo): Handle cgraph_remove_functions.
2661 2007-01-08  Nick Clifton  <nickc@redhat.com>
2663         * config/frv/predicates.md (reg_or_0_operand): Accept
2664         CONST_DOUBLEs.
2666 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
2668         * config/bfin/rtems.h, config/bfin/t-rtems: New.
2669         * config.gcc: Add bfin*-rtems*.
2671 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2673         * c.opt: Add -flax-vector-conversions.
2674         * c-typeck.c (convert_for_assignment): Pass flag to
2675         vector_types_convertible_p to allow emission of note.
2676         (digest_init): Likewise.
2677         * c-opts.c: Handle -flax-vector-conversions.
2678         * c-common.c (flag_lax_vector_conversions): New.
2679         (vector_types_convertible_p): Unless -flax-vector conversions
2680         has been passed, disallow conversions between vectors with
2681         differing numbers of subparts and/or element types.  If such
2682         a conversion is disallowed, possibly emit a note on the first
2683         occasion only to inform the user of -flax-vector-conversions.
2684         The new last argument specifies this.
2685         * c-common.h (flag_lax_vector_conversions): New.
2686         (vector_types_convertible_p): Add extra argument.
2687         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2688         char_type_node for V*QI type vectors.
2689         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2690         Update to satisfy new typechecking rules.
2691         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2692         C and C++ variants.
2693         * doc/invoke.texi (C Dialect Options): Document
2694         -flax-vector-conversions.
2696 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2698         PR tree-optimization/29877
2699         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2700         a register variable on the RHS to not be replaceable.
2702 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
2703         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2704         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2705         const_simm15, const_pow2, const_npow2): Added.
2706         * config/score/misc.md (insv, extv, extzv, movmemsi,
2707         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2708         * config/score/score.c (score_address_cost, score_select_cc_mode):
2709         Added.
2710         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2711         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2712         Update score_print_operand.
2713         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2714         Adjust register allocate order and update some macro define.
2715         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2716         mdx_block_move_straight, mdx_block_move_loop_head,
2717         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2718         mdx_block_move): Added.
2719         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2720         mdp_limm): Updated and fix some bug and typo.
2721         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2722         (movsf, movdf, doloop_end): Added.
2724 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
2726         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2727         config/arm/thumb2.md: Fix comment typos.
2728         * doc/extend.texi: Fix a typo.
2730 2007-01-07  Eric Christopher  <echristo@apple.com>
2732         * configure.ac: Check for __stack_chk_fail for darwin.
2733         * configure: Regenerate.
2735 2007-01-07  Richard Guenther  <rguenther@suse.de>
2737         * tree-vrp.c (extract_range_from_assert): CSE calls to
2738         compare_values where possible.
2739         (extract_range_from_unary_expr): Likewise.
2741 2007-01-07  Anatoly Sokolov <aesok@post.ru>
2743         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2744         * config/avr/avr.c (call_insn_operand): Delete function.
2745         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2746         instead of register number. Use predicates.
2747         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2748         call_insn_operand): Add.
2750 2007-01-06  Jan Hubicka  <jh@suse.cz>
2752         * tree-pass.h (pass_build_cgraph_edges): Declare.
2753         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2754         (visited_nodes): Remove.
2755         (cgraph_create_edges): Move to cgraphbuild.c; rename to
2756         build_cgrpah_edges; make visited_nodes local.
2757         (cgraph_process_new_functions): DO not call initialize_inline_failed.
2758         (record_references_in_initializer): Move to cgraphbuild.c
2759         (initialize_inline_failed, rebuild_cgraph_edges,
2760         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2761         (verify_cgraph_node): Make visited_nodes local.
2762         (cgraph_analyze_function): Do not call cgraph_create_edges and
2763         initialize_inline_failed.
2764         (cgraph_expand_function): Do not call cgraph_lower_function;
2765         assert that function is already lowered.
2766         * Makefile.in (cgraphbuild.o): New.
2767         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2768         at the end of lowering passes.
2770 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
2772         * ifcvt.c (cond_move_convert_if_block): New function, code
2773         factored out from...
2774         (cond_move_process_if_block): ...here.  Call the new function
2775         on the THEN and ELSE blocks.
2776         (merge_if_block): Do not copy global_live_at_end, merge_blocks
2777         already takes care of this.
2779 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2781         PR c/19978
2782         * tree.h (TREE_OVERFLOW_P): New.
2783         * c-typeck.c (parser_build_unary_op): Warn only if result
2784         overflowed and operands did not.
2785         (parser_build_binary_op): Likewise.
2786         (convert_for_assignment): Remove redundant overflow_warning.
2787         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2789 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2791         * c-typeck.c (store_init_value): Split over two lines to follow
2792         the GNU coding style.
2794 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
2796         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2797         __GXX_EXPERIMENTAL_CXX0X__.
2798         * doc/cpp.texi: Same.
2800 2007-01-05  Richard Guenther  <rguenther@suse.de>
2802         PR middle-end/27826
2803         * tree.c (get_narrower): Do not construct COMPONENT_REFs
2804         with mismatched types.  Instead explicitly build a
2805         conversion NOP_EXPR.
2807 2007-01-05  Ian Lance Taylor  <iant@google.com>
2809         * c-common.c (decl_with_nonnull_addr_p): New function.
2810         (c_common_truthvalue_conversion): Call it.
2811         * c-typeck.c (build_binary_op): Likewise.
2812         * c-common.h (decl_with_nonnull_addr_p): Declare.
2814 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
2816         PR c/30360
2817         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2818         denom against 0.0.
2820 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
2822         * doc/install.texi (Final install): Document the fact that
2823         the GNAT runtime should not be stripped.
2825 2007-01-04  Jan Hubicka  <jh@suse.cz>
2827         * tree-inline.c (fold_marked_statements): Update operand caches
2828         and EH after folding
2830 2007-01-04  Ian Lance Taylor  <iant@google.com>
2832         * c-common.c (check_function_nonnull): Whitespace fix.
2834 2007-01-04  Jan Hubicka  <jh@suse.cz>
2836         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2837         comitted older version of patch.
2838         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2839         function, celanup cfg and collect garbage.
2841 2007-01-04  Mike Stump  <mrs@apple.com>
2843         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2844         (clean): Likewise.
2846 2007-01-04  Eric Christopher  <echristo@apple.com>
2848         * libgcc2.c (__bswapsi2): Use SItype.
2849         (__bswapdi2): Use DItype.
2850         * libgcc2.h: Update for above.
2852 2007-01-04  Paul Brook  <paul@codesourcery.com>
2854         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2855         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2856         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2857         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2858         umulsi3_highpart): Make conditional on !arm_arch6.
2859         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2860         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2861         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2862         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2863         umulsi3_highpart_v6): New insns.
2865 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2867         * fold-const.c (fold_convert): When casting an expression to void,
2868         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2869         have a type.  Instead of attempting to build a NOP_EXPR, return
2870         these "special" trees directly.
2872 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2874         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2875         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2876         rs6000_cost->fp not 0 as adjustment for outer NEG.
2878 2007-01-04  Jan Hubicka  <jh@suse.cz>
2880         * cgraph.c (cgraph_release_function_body): New function.
2881         (cgraph_remove_node): Use it.
2882         * cgraph.h (cgraph_release_function_body): Declare.
2883         * cgraphunit.c (cgraph_expand_function): Use it.
2884         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2885         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2886         is initialized and while compilation of other function is running.
2887         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2888         statement CFG annotations from here to ...
2889         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2891 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2893         * cfgloop.h (enum li_flags): Make the constants powers of two.
2895 2007-01-04  Jan Hubicka  <jh@suse.cz>
2897         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2898         set.
2899         (fold_marked_statements): New function.
2900         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2901         * tree-inline.h (copy_body_data): Add statements_to_fold.
2903 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2905         * config.gcc: Mention libgcc/config.host.
2906         * Makefile.in: Update comments mentioning libgcc.
2907         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2908         (all.cross, start.encap, rest.encap, rest.cross): Update
2909         dependencies for libgcc move.
2910         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2911         (clean-target, clean-target-libgcc): Delete.
2912         (srcdirify, GCC_EXTRA_PARTS): New macros.
2913         (libgcc-support, libgcc.mvars): New rules.
2914         (distclean): Remove mention of mklibgcc.
2915         (install): Don't reference INSTALL_LIBGCC.
2916         (install-common): Don't reference EXTRA_PARTS.
2917         (install-libgcc, install-multilib): Delete rules.
2918         * mklibgcc.in: Delete file.
2919         * doc/configfiles.texi: Don't mention mklibgcc.
2921         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2922         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2923         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2924         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2925         (mklibgcc, ldblspecs): Likewise.
2927         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2928         (posixpre.def): Use $(T).
2929         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2930         (SHLIB_SRC, SHLIB_INSTALL): Delete.
2931         (SHLIB_LINK): Make dummy.
2932         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2934         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2936         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2937         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2938         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2939         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2941         * configure.ac (all_outputs): Remove mklibgcc.
2942         * configure: Regenerated.
2944 2007-01-03  Josh Conner  <jconner@apple.com>
2946         PR middle-end/29683
2947         * calls.c (compute_argument_addresses): Set stack and stack_slot
2948         for partial args, too.
2949         (store_one_arg): Use locate.size.constant for the size when
2950         generating a save_area.
2952 2007-01-03  Robert Kennedy <jimbob@google.com>
2954         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2955         whose definitions are deleted due to basic block merging.
2957 2007-01-03  Paul Brook  <paul@codesourcery.com>
2959         PR target/16634
2960         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2961         functions.
2962         (use_return_insn): Return 0 for Thumb interrupt functions.
2963         (print_multi_reg): Add rfe argument for IRQ returns.
2964         (arm_output_epilogue): Pop interrupt return address directly into PC.
2965         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2967 2007-01-03  Paul Brook  <paul@codesourcery.com>
2969         Merge from sourcerygxx-4_1.
2970         * config/arm/thumb2.md: New file.
2971         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2972         Thumb-2.
2973         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2974         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2975         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2976         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2977         tables.
2978         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2979         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2980         comptibility.
2981         * config/arm/ieee754-sf.S: Ditto.
2982         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2983         (thumb1_base_register_rtx_p): ... to this.
2984         (thumb_index_register_rtx_p): Rename...
2985         (thumb1_index_register_rtx_p): ... to this.
2986         (thumb_output_function_prologue): Rename...
2987         (thumb1_output_function_prologue): ... to this.
2988         (thumb_legitimate_address_p): Rename...
2989         (thumb1_legitimate_address_p): ... to this.
2990         (thumb_rtx_costs): Rename...
2991         (thumb1_rtx_costs): ... to this.
2992         (thumb_compute_save_reg_mask): Rename...
2993         (thumb1_compute_save_reg_mask): ... to this.
2994         (thumb_final_prescan_insn): Rename...
2995         (thumb1_final_prescan_insn): ... to this.
2996         (thumb_expand_epilogue): Rename...
2997         (thumb1_expand_epilogue): ... to this.
2998         (arm_unwind_emit_stm): Rename...
2999         (arm_unwind_emit_sequence): ... to this.
3000         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3001         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3002         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3003         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3004         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3005         thumb2_output_casesi): New functions.
3006         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3007         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3008         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3009         THUMB2_WORK_REGS): Define.
3010         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3011         arm_condexec_mask, arm_condexec_masklen)): New variables.
3012         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3013         (arm_override_options): Check new CPU capabilities.
3014         Set new architecture flag variables.
3015         (arm_isr_value): Handle v7m interrupt functions.
3016         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
3017         Thumb-2.
3018         (const_ok_for_arm): Handle Thumb-2 constants.
3019         (arm_gen_constant): Ditto.  Use movw when available.
3020         (arm_function_ok_for_sibcall): Return false for v7m interrupt
3021         functions.
3022         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3023         (thumb_find_work_register, arm_load_pic_register,
3024         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3025         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3026         print_multi_reg, output_mov_long_double_fpa_from_arm,
3027         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3028         output_mov_double_fpa_from_arm, output_move_double,
3029         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3030         output_return_instruction, arm_output_function_prologue,
3031         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3032         arm_output_mi_thunk, thumb_set_return_address): Ditto.
3033         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
3034         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3035         (arithmetic_instr, shift_op): Use arm_shift_nmem.
3036         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
3037         '!' and 'L'.
3038         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3039         (thumb_expand_prologue): Use thumb_set_frame_pointer.
3040         (arm_file_start): Output directive for unified syntax.
3041         (arm_unwind_emit_set): Handle stack alignment instruction.
3042         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3043         Add v6t2, v7, v7a, v7r and v7m.
3044         (RETLDM): Add Thumb-2 code.
3045         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3046         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3047         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3048         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3049         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3050         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3051         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3052         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3053         ADJUST_INSN_LENGTH): Define.
3054         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3055         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3056         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3057         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3058         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3059         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3060         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3061         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3062         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3063         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3064         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3065         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3066         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3067         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3068         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3069         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3070         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3071         cortex-m3.
3072         * config/arm/arm-tune.md: Regenerate.
3073         * config/arm/arm-protos.h: Update prototypes.
3074         * config/arm/vfp.md: Enable patterns for Thumb-2.
3075         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
3076         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3077         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3078         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3079         * config/arm/libunwind.S: Add Thumb-2 code.
3080         * config/arm/constraints.md: Update include Thumb-2.
3081         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3082         * config/arm/ieee754-df.S: Ditto.
3083         * config/arm/bpabi.S: Ditto.
3084         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3085         * config/arm/predicates.md (low_register_operand,
3086         low_reg_or_int_operand, thumb_16bit_operator): New.
3087         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3088         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3089         * config/arm/t-arm-elf: Add armv7 multilib.
3090         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3091         Include thumb2.md.
3092         (UNSPEC_STACK_ALIGN, ce_count): New.
3093         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3094         arm_zero_extendsidi2, arm_zero_extendqidi2): New
3095         insns/expanders.
3096         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3097         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3098         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3099         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3100         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3101         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3102         insns.
3103         * doc/extend.texi: Document ARMv7-M interrupt functions.
3104         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3106 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
3108         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3109         (struct _Unwind_Context): Rename args_size to flags, remove
3110         signal_frame field, add a new args_size field and version field.
3111         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3112         _Unwind_IsExtendedContext): New inline functions.
3113         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3114         Assume by_value array is only present if _Unwind_IsExtendedContext.
3115         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3116         _Unwind_IsSignalFrame.
3117         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3118         (uw_update_context_1): Use _Unwind_SetSignalFrame.
3119         (uw_init_context_1): Initialize context->flags to
3120         EXTENDED_CONTEXT_BIT.
3121         * config/rs6000/linux-unwind.h (frob_update_context): Use
3122         _Unwind_SetSignalFrame.
3124 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3126         PR middle-end/30353
3127         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3128         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3129         the value.
3131 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3133         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3134         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3135         RET, AT_SP): Remove.
3136         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3137         instead of RET.
3139         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3140         * cfgbuild.c (find_basic_blocks): Likewise.
3141         * cfgrtl.c (rtl_create_basic_block): Likewise.
3142         * function.c (temp_slots_at_level): Likewise.
3143         * reg-stack.c (stack_regs_mentioned): Likewise.
3144         * regclass.c (allocate_reg_info): Likewise.
3145         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3146         set_bb_for_stmt, move_block_to_fn): Likewise.
3147         * tree-complex.c (tree_lower_complex): Likewise.
3148         * vec.h (VEC_safe_grow_cleared): New.
3150         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3152 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
3154         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3155         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3156         to loop_version.
3157         * cfgloopmanip.c (scale_loop_frequencies): Export.
3158         (loopify): Scale the frequencies by prescribed coefficients.
3159         (set_zero_probability): New function.
3160         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3161         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3162         and frequencies according to arguments.
3163         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3164         correctly.
3165         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3166         * modulo-sched.c (sms_schedule): Set probabilities for entering
3167         versioned loop correctly.
3168         * tree-vect-transform.c (vect_transform_loop): Ditto.
3169         * cfgloop.h (loopify, loop_version): Declaration changed.
3170         (scale_loop_frequencies): Declared.
3172 2007-01-02  Jan Hubicka  <jh@suse.cz>
3174         * cgraph.c: Include tree-flow.h
3175         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3176         early_local_passes.
3177         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3178         * tree-pass.h (pass_all_early_optimizations): Declare.
3179         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3180         early_local_passes.
3181         (cgraph_analyze_function): Do early_local_passes.
3182         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3183         Do not add referenced vars.
3184         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3185         SSA form.
3186         (gate_all_early_local_passes): New gate.
3187         (pass_early_local_passes): Use new gate.
3188         (execute_early_local_optimizations): New functions.
3189         (gate_all_early_optimizations): New gate.
3190         (pass_all_early_optimizations): New pass.
3191         (execute_free_datastructures): Free SSA only when initialized.
3192         (gate_init_datastructures): Init only when optimizing.
3193         (tree_lowering_passes): Do early local passes when called late.
3194         * tree-profile.c (do_tree_profiling): Don't profile functions added
3195         late.
3196         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3197         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3198         not active.
3199         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3200         during early_local_passes.
3201         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3204 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3206         * Makefile.in: Update copyright year.
3208 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3210         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3211         in $(lang_checks) and check-consistency targets.
3213 2007-01-02  Jan Hubicka  <jh@suse.cz>
3215         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3216         Do not add referenced vars.
3217         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3218         are not active.
3219         * passes.c (init_optimization_passes): Put mudflap_2 after
3220         free_datastructures.
3222 2007-01-02  Jan Hubicka  <jh@suse.cz>
3224         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3225         Set NOTHROW flag on call statements proved to be nothrow.
3226         Update statement of local calls so new pure/const functions are
3227         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3228         goto receivers.
3229         (tree_rest_of_compilation): Register hooks and initialize bitmap
3230         early. Do not set after_inlining flag.
3232 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3234         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3235         HOST_BITS_PER_LONGLONG
3237 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3239         PR c/19977
3240         * c-typeck.c (store_init_value): Don't emit pedantic overflow
3241         warning for non-static initializers.
3243 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
3245         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3246         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3247         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3248         Always use set_unique_reg_note to add REG_EQUAL notes.
3250 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3252         Revert:
3253         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3255         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3256         * cfgbuild.c (find_basic_blocks): Likewise.
3257         * cfgrtl.c (rtl_create_basic_block): Likewise.
3258         * function.c (temp_slots_at_level): Likewise.
3259         * reg-stack.c (stack_regs_mentioned): Likewise.
3260         * regclass.c (allocate_reg_info): Likewise.
3261         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3262         set_bb_for_stmt, move_block_to_fn): Likewise.
3263         * tree-complex.c (tree_lower_complex): Likewise.
3264         * vec.h (VEC_safe_grow_cleared): New.
3266 2007-01-02  Ian Lance Taylor  <iant@google.com>
3268         * c-common.c (c_common_truthvalue_conversion): When warning about
3269         using an assignment as a truth value, set TREE_NO_WARNING.
3271 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3273         PR middle-end/7651
3274         * c.opt (Wold-style-declaration): New.
3275         * doc/invoke.texi (C-only Warning Options): New.
3276         (Wold-style-declaration): Document it.
3277         (Wextra): Enabled by -Wextra.
3278         * c-opts.c (c_common_post_options): Enabled by -Wextra.
3279         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3280         -Wold-style-declaration.
3282 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3284         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3285         * cfgbuild.c (find_basic_blocks): Likewise.
3286         * cfgrtl.c (rtl_create_basic_block): Likewise.
3287         * function.c (temp_slots_at_level): Likewise.
3288         * reg-stack.c (stack_regs_mentioned): Likewise.
3289         * regclass.c (allocate_reg_info): Likewise.
3290         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3291         set_bb_for_stmt, move_block_to_fn): Likewise.
3292         * tree-complex.c (tree_lower_complex): Likewise.
3293         * vec.h (VEC_safe_grow_cleared): New.
3295 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
3297         * c-common.c (c_common_nodes_and_builtins): Since variants of
3298         void_type_node get built before it is given a name, we need to
3299         give those variants the name, too.
3300         (complete_array_type): We need to work with the canonical main
3301         type of the array, from which we will build the qualified version.
3302         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3303         * print-tree.c (print_node): Display canonical type information
3304         for each type.
3305         * stor-layout.c (layout_type): When we don't know the
3306         alignment of a type for which we're building an array, we end up
3307         guessing wrong, so make the type require structural equality.
3308         * tree.c (make_node_stat): When we build a new type, it is its
3309         own canonical type.
3310         (build_type_attribute_qual_variant): When building an attribute
3311         variant, its canonical type is the non-attribute variant. However,
3312         if the attributes are target-dependent and they differ, we need to
3313         use structural equality checks for this type.
3314         (build_qualified_type): A qualified type is not equivalent to its
3315         unqualified variant; set the canonical type appropriately.
3316         (build_distinct_type_copy): When building a distinct type from
3317         another type, the new type is its own canonical type.
3318         (build_variant_type_copy): When building a new type variant, we
3319         assume that it is equivalent to the original type.
3320         (build_pointer_type_for_mode): When building a pointer type, also
3321         build a canonical type pointer.
3322         (build_reference_type_for_mode): When building a reference type,
3323         also build a canonical type reference.
3324         (build_index_type): When we can't hash an index type (e.g.,
3325         because its maximum value is negative), the index type requires
3326         structural equality tests.
3327         (build_array_type): Build the canonical form of an array type.
3328         (build_function_type): Function types require structural equality,
3329         because they contain default arguments, attributes, etc.
3330         (build_method_type_directly): Ditto for method types.
3331         (build_offset_type): Build the canonical offset type.
3332         (build_complex_type): Build the canonical vector type.
3333         (make_vector_type): Build the canonical vector type.
3334         * tree.h (TYPE_CANONICAL): New.
3335         (TYPE_STRUCTURAL_EQUALITY_P): New.
3336         (SET_TYPE_STRUCTURAL_EQUALITY): New.
3337         (struct tree_type): Added "canonical" field.
3338         * params.h (VERIFY_CANONICAL_TYPES): New.
3339         * doc/c-tree.texi (TYPE_CANONICAL): Document.
3340         (TYPE_STRUCTURAL_EQUALITY_P): Document.
3341         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3342         * doc/invoke.texi (verify-canonical-types): Document --param
3343         parameter for verifying canonical types.
3345 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3347         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3348         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3349         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3350         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3351         tm_file.
3352         * config/rs6000/e500.h: New.
3353         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3354         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3355         Remove.
3356         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3357         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3358         TARGET_E500_DOUBLE): Remove.
3359         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3360         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3361         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3362         * config/rs6000/rs6000.c (rs6000_override_options): Use
3363         CHECK_E500_OPTIONS.
3365 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3367         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3368         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3369         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3370         rs6000_emit_cmove): Don't check TARGET_E500.
3371         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3372         sordered): Don't check TARGET_E500.
3374 2007-01-01  Eric Christopher  <echristo@apple.com>
3376         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3377         HARD_REGISTER_NUM_P.
3379 2007-01-01  Roger Sayle  <roger@eyesopen.com>
3381         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3382         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3383         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3384         eq/ne Y".
3386 2007-01-01  Mike Stump  <mrs@apple.com>
3388         * configure.ac: Remove support for building with Apple's gcc-3.1.
3390 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3392         PR middle-end/30311
3393         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3394         REG is a hard register.  Do not modify REG before calling
3395         subreg_nregs.
3396         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3397         multiple of the size of YMODE for certain lowpart cases.
3399 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3401         PR middle-end/30253
3402         * gimplify (voidify_wrapper_expr): Update for
3403         GIMPLE_MODIFY_STMT.
3405 2007-01-01  Andreas Schwab  <schwab@suse.de>
3407         PR target/29166
3408         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3409         save of BR0 in extra_spill_size instead of spill_size.
3410         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3411         area.
3412         (ia64_expand_epilogue): Restore BR0 from its new location.
3414 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3416         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3417         Use a temporary variable if the left hand side is not a gimple
3418         register.
3420 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3422         * gimplify.c (gimplify_return_expr): Make the temporary variable
3423         for the return expression, a gimple register variable.
3425 2007-01-01  Jan Hubicka  <jh@suse.cz>
3427         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3428         INSN_LIST.
3430 2007-01-01  Mike Stump  <mrs@apple.com>
3432         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3433         .literal16.
3434         * config/darwin.c (machopic_select_rtx_section): Use
3435         HAVE_GAS_LITERAL16.
3436         (darwin_mergeable_constant_section): Likewise.
3437         * configure: Regenerate.
3438         * config.in: Regenerate.
3440 2007-01-01  Jan Hubicka  <jh@suse.cz>
3441             Andrew Pinski  <pinskia@gmail.com>
3443         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3444         before starting IPA passes.