* cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
[official-gcc.git] / gcc / ChangeLog
blob64e97e50c5bd82cf9b460366027c1b58a306f309
1 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
3         * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
4         ipa-inline.c, tree-profile.c, tree-ssa-live.c,
5         tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
6         value-prof.c: Fix comment typos.
8 2007-01-30  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
10         PR c++/24745
11         * doc/invoke.texi (Wpointer-arith): Document warning.
12         
13 2007-01-30  Janis Johnson
15         * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
16         the list of discrepancies from the draft TR.
18 2007-01-30  Dirk Mueller  <dmueller@suse.de>
20         PR c++/30601
21         * doc/invoke.texi (-Wreturn-type): Update description to
22         match new behavior. 
24 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
26         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
28 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
30         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
31         to handle optabs that operate on floating point input argument and
32         output to integer output.
33         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
34         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
35         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
36         expand_builtin_interclass_mathfn ().
37         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
38         (ilogbsi2): Remove.
39         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
40         ilogbl built-in functions as x87 intrinsics.
42 2007-01-30  Richard Guenther  <rguenther@suse.de>
44         PR middle-end/27657
45         * dwarf2out.c (reference_to_unused): Query varpool if the
46         variable was output.
48 2007-01-30  Richard Guenther  <rguenther@suse.de>
50         PR middle-end/30313
51         * passes.c (execute_one_pass): Reset in_gimple_form to not
52         confuse non-unit-at-a-time mode.
54 2007-01-29  Roger Sayle  <roger@eyesopen.com>
55             Richard Guenther  <rguenther@suse.de>
57         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
58         avoid undefined behaviour on overflow.  Use force_fit_type_double
59         to construct the constant with the specified TREE_OVERFLOW.
61 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
63         * config/dfp-bit.c: Add parameterized support for fp exceptions.
64         * config/dfp-bit.h: Ditto.
66 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
68         * c-decl.c (pop_scope): Replace warnings with call to
69         warn_for_unused_label.
70         * c-common.h (warn_for_unused_label): Declare.
71         * c-common.c (warn_for_unused_label): Define.
73 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
75         * tree-optimize.c (update_inlined_to_pointers): Delete unused
76         function.
78 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
80         * Makefile.in (USER_H): Remove decfloat.h.
81         * ginclude/decfloat.h: Delete, moving contents to ...
82         * ginclude/float.h: Add support for decimal floating point,
83         guarded by __STDC_WANT_DEC_FP__.
85 2007-01-29  Mike Stump  <mrs@apple.com>
87         * doc/gccint.texi (Top): Rename Loop Representation to Loop
88         Analysis and Representation to resolve case insensitive conflict.
89         * doc/loop.texi (Loop Analysis and Representation): Likewise.
91 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
93         * tree.h (struct tree_memory_tag): Add aliases member.
94         (MTAG_ALIASES): New macro.
95         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
96         (add_may_alias): Remove pointer-set. Update for may_aliases being
97         a bitmap. 
98         (mark_aliases_call_clobbered): Update for may_aliases being a
99         bitmap.
100         (compute_tag_properties): Ditto.
101         (create_partition_for): Ditto.
102         (compute_memory_partitions): Ditto.
103         (dump_may_aliases_for): Ditto.
104         (is_aliased_with): Ditto.
105         (add_may_alias_for_new_tag): Ditto.
106         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
107         (compute_is_aliased): New function.
108         (compute_may_aliases): Call compute_is_aliased).
109         (init_alias_info): Initialize alias_bitmap_obstack.
110         (union_alias_set_into): New function.
111         (compute_flow_sensitive_aliasing): Use union_aliases_into.
112         (have_common_aliases_p): Rewrite to take two bitmaps and use
113         intersection.
114         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
115         Update for bitmaps.
116         (finalize_ref_all_pointers): Update for add_may_alias changes.
117         (new_type_alias): Ditto.
118         * tree-flow-inline.h (may_aliases): Return a bitmap.
119         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
120         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
121         may_aliases being a bitmap.
122         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
123         may_aliases now returns a bitmap.
124         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
125         being a bitmap.
126         * tree-ssa-operands.c (add_virtual_operand): Update for
127         may_aliases being a bitmap.
128         
129 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
131         PR tree-optimization/30630
132         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
133         variable as changed.
135 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
137         PR c++/28266
138         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
139         expanded only once even if an error occurs.
141 2007-01-29  Ben Elliston  <bje@au.ibm.com>
143         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
144         (__gcov_average_profiler): Likewise.
145         (__gcov_ior_profiler): Likewise.
146         (__gcov_merge_ior): Likewise.
148 2007-01-28  Jan Hubicka  <jh@suse.cz>
150         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
151         * value-prof.c (stringop_block_profile): Likewise.
153 2007-01-28  Jan Hubicka  <jh@suse.cz>
155         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
156         variant handling histograms; add wrapper.
157         (clear_storage_via_libcall): Export.
158         (emit_block_move_hints): Break out from ...; add histograms.
159         (emit_block_move): ... this one.
160         (clear_storage_hints): Break out from ...; add histograms.
161         (clear_storage): ... this one.
162         (set_storage_via_memset): Handle histogram.
163         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
164         (clear_storage_hints, clear_storage_via_libcall): Declare.
165         (set_storage_via_setmem): Update prototype.
166         * doc/md.texi (movmem, setmem): Document new arguments.
168         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
169         new histograms.
170         (stringop_block_profile): New global function.
171         (tree_stringops_values_to_profile): Profile block size and alignment.
172         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
173         HIST_TYPE_IOR.
174         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
175         (stringop_block_profile): Declare.
176         * builtins.c: Include value-prof.h.
177         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
178         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
179         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
180         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
181         * profile.c (instrument_values): Add new counters.
182         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
183         calls.
184         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
185         (tree_init_edge_profiler): Build new profilers.
186         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
187         (pass_tree_profile): Add dump.
188         (tree_profile_hooks): Update.
189         * Makefile.in (LIBGCOV): Add new constants.
190         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
191         __gcov_ior_profiler): New.
192         * i386.md (movmem/setmem expanders): Add new optional arguments.
194 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
196         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
198 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
200         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
201         (tree_unswitch_single_loop): Pass eni_size_weights to
202         tree_num_loop_insns.
203         * tree-ssa-loop-manip.c: Include tree-inline.h.
204         (can_unroll_loop_p): Pass eni_size_weights to
205         tree_num_loop_insns.
206         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
207         Pass eni_size_weights to estimate_num_insns.
208         * tree.h (init_inline_once): Export.
209         * toplev.c (backend_init): Call init_inline_once.
210         * cgraphunit.c (cgraph_process_new_functions,
211         cgraph_analyze_function): Pass eni_inlining_weights to
212         estimate_num_insns.
213         * ipa-inline.c (compute_inline_parameters): Ditto.
214         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
215         to estimate_num_insns.
216         (try_unroll_loop_completely): Pass eni_size_weights to
217         tree_num_loop_insns.
218         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
219         ot estimate_num_insns.
220         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
221         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
222         * tree-inline.c (eni_inlining_weights, eni_size_weights,
223         eni_time_weights): New variables.
224         (init_inline_once): Initialize them.
225         (struct eni_data): Mew.
226         (estimate_num_insns_1, estimate_num_insns): Use weights.
227         * tree-inline.h (struct eni_weights_d): New.
228         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
229         (estimate_num_insns): Declaration changed.
230         * cfgloop.h (tree_num_loop_insns): Declaration changed.
231         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
232         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
234 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
236         * tree-data-ref.c (conflict_fn): Assert that the number of affine
237         relations in the conflict function is valid.
239 2007-01-27  Ian Lance Taylor  <iant@google.com>
241         * common.opt: Add fstrict-overflow.
242         * opts.c (decode_options): Set flag_strict_overflow if -O2.
243         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
244         (TYPE_OVERFLOW_UNDEFINED): Define.
245         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
246         Replace all uses.
247         * tree.h (TYPE_TRAP_SIGNED): Don't define.
248         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
249         (fold_negate_expr): Likewise.
250         (make_range): Likewise.
251         (extract_muldiv_1): Likewise.
252         (maybe_canonicalize_comparison): Likewise.
253         (fold_comparison): Likewise.
254         (fold_binary): Likewise.
255         (tree_expr_nonnegative_p): Likewise.
256         (tree_expr_nonzero_p): Likewise.
257         * tree-vrp.c (compare_values): Likewise.
258         (extract_range_from_binary_expr): Likewise.
259         (extract_range_from_unary_expr): Likewise.
260         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
261         Likewise.
262         (nowrap_type_p): Likewise.
263         * tree-scalar-evolution.c (simple_iv): Likewise.
264         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
265         (build_range_check): Likewise.
266         (extract_muldiv_1): Likewise.
267         (fold_comparison): Likewise.
268         * tree-vrp.c (vrp_int_const_binop): Likewise.
269         (extract_range_from_unary_expr): Likewise.
270         * convert.c (convert_to_integer): Likewise.
271         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
272         (fold_comparison): Likewise.
273         (fold_binary): Likewise.
274         * optabs.c (optab_for_tree_code): Likewise.
275         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
276         * simplify-rtx.c (simplify_const_relational_operation): Check
277         flag_strict_overflow and flag_trapv.
278         (simplify_const_relational_operation): Likewise.
279         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
280         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
281         -fstrict-overflow.
283 2007-01-27  Roger Sayle  <roger@eyesopen.com>
285         * tree.c (tree_fold_gcd): Delete.
286         * tree.h (tree_fold_gcd): Remove prototype.
287         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
288         test whether one constant integer is a multiple of another.  Instead
289         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
290         * fold-const.c (multiple_of_p):  We've determined both TOP and
291         BOTTOM are integer constants so we can call int_const_binop directly
292         instead of the more generic const_binop.
294 2007-01-27  Roger Sayle  <roger@eyesopen.com>
296         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
297         1*X check that the constant hasn't overflowed, to preserve the
298         TREE_OVERFLOW bit.
299         (round_up): Provide an efficient implementation when rouding-up an
300         INTEGER_CST to a power-of-two.
302 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
304         * doc/sourcebuild.texi: Add comma for clarity.
305         * doc/extend.texi: Fix some typos.
306         * doc/passes.texi: Likewise.
307         * doc/cppinternals.texi: Likewise.
308         * doc/c-tree.texi: Likewise.
309         * doc/tree-ssa.texi: Likewise.
310         * doc/install.texi: Likewise.
312 2007-01-27  Jan Hubicka  <jh@suse.cz>
314         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
315         (pass_sra): Do not require alias information.
316         * passes.c (init_optimization_passes): Add SRA
318 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
320         * tracer.c (rest_of_handle_tracer): We already cleaned
321         up the CFG in tracer() so don't do it here again.
322         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
323         cleanup_cfg here, either.  And don't call renumber_insns.
325         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
326         and BB_HEAD are in the insn stream here.  Instead make sure
327         that BB_INSN is valid on all insns.  Also, do check here that
328         there are no pending branch predictions...
329         (rtl_verify_flow_info): ...instead of doing it here.  Checks
330         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
331         here.
333 2007-01-26  Roger Sayle  <roger@eyesopen.com>
335         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
336         function to simplify/factorize operand order canonicalization.
337         (ix86_fixup_binary_operands): Reorganize using the above function.
338         (ix86_binary_operator_ok): Likewise.
340 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
342         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
343         file scope from optimize_attrs.
344         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
345         up search.
346         (optimize_attrs): Clear insn_code_values after freeing it.
348 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
350         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
351         a parts.base assignment.
353 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
355         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
356         (compute_subscript_distance, initialize_data_dependence_relation,
357         finalize_ddr_dependent, analyze_ziv_subscript,
358         analyze_siv_subscript_cst_affine,
359         compute_overlap_steps_for_affine_univar,
360         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
361         analyze_siv_subscript, analyze_miv_subscript,
362         analyze_overlapping_iterations, subscript_dependence_tester_1,
363         compute_self_dependence, free_dependence_relation): Work
364         with affine_fn instead of chrecs.
365         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
366         common_affine_function, affine_function_base,
367         affine_function_constant_p, affine_fn_op, affine_fn_plus,
368         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
369         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
370         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
371         (all_chrecs_equal_p): Removed.
372         * tree-data-ref.h (affine_fn, conflict_function): New types.
373         (struct subscript): Change type of conflicting_iterations_in_a
374         and conflicting_iterations_in_b.
376 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
378         PR other/30182
379         * config/pa/pa.h (TARGET_HPUX_11): New.
380         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
381         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
383 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
384         Richard Guenther  <rguenther@suse.de>
386         * tree-ssa-structalias.c (solve_graph): Handle case
387         we merged the variable to another.
389 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
391         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
392         functions.
394         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
395         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
397 2007-01-25  DJ Delorie  <dj@redhat.com>
399         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
400         allow changes to modes which don't fit in those registers.
402         * reload1.c (choose_reload_regs): Check for invalid subregs before
403         computing their locations, not after.
405 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
407         PR 25127
408         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
409         save Altivec registers in an eh_return function.
410         (compute_vrsave_mask): Likewise.
411         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
412         for saving Altivec registers.
413         (rs6000_emit_prologue): Don't allocate stack twice in
414         eh_return function.  Correct expected value of altivec_save_offset
415         when using save_world.  Describe save of R0 to stack when using
416         save_world.  Describe stack pointer adjustment when using
417         save_world.  Remove duplicated eh_return parameter register saving.
418         Update sp_offset variable after save_world.
419         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
420         darwin-world.asm.
421         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
422         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
423         implies Altivec.
425 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
427         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
429 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
431         PR other/30182
432         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
433         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
435 2007-01-25  Richard Guenther  <rguenther@suse.de>
437         * doc/invoke.texi (-Wcoverage-mismatch): Document.
438         * common.opt (-Wcoverage-mismatch): New warning option.
439         * coverage.c (get_coverage_counts): Ignore coverage mismatch
440         if -Wcoverage-mismatch is given.
442 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
444         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA. 
445         (ipcp_driver): Change to static definition. 
446         Add dumping of the ifunctions.
447         (constant_val_insert): Remove unused parameter. Support for SSA.
448         (ipcp_propagate_const): Support for SSA.
449         (ipcp_profile_bb_print): Print only analyzed nodes.
450         (ipcp_replace_map_create): Remove support for Fortran constant 
451         for now.
452         * ipa-prop.c (ipa_method_modify_stmt, 
453         ipa_callsite_compute_param): Support for SSA.
454         * ipa-prop.h (ipcp_driver): Remove declaration.
455         (IS_VALID_TREE_MAP_INDEX): Add define.
457 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
459         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
461 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
463         PR middle-end/30447
464         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
465         constant argument to cabs and do it without checking for
466         -funsafe-math-optimizations.
468 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
470         * c-common.h (RID_FIRST_CXX0X): New.
471         (RID_LAST_CXX0X): New.
472         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
473         by -Wall.
474         * c.opt (Wc++0x-compat): New.
475         * doc/invoke.texi (-Wc++0x-compat): Document.
477 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
479         * builtins.c (fold_builtin_carg): New.
480         (fold_builtin_1): Use it.
482 2007-01-24  Jan Hubicka  <jh@suse.cz>
484         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
486         * ipa-inline.c (initial_insns, max_insns): Delete.
487         (compute_max_insns): New function.
488         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
489         of insns as base for code growth.
490         (cgraph_decide_inlining): Make initial_insns local; do not compute
491         max_insns.
492         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
493         * doc/invoke.texi (inline-unit-growth): Update docs.
495 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
497         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
498         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
499         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
500         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
501         for CPUs that have PTA_CX16 set.
503         PR middle-end/27416
504         * gimplify.c (omp_check_private): New function.
505         (gimplify_scan_omp_clauses): Use it for
506         firstprivate/lastprivate/reduction.
508         PR middle-end/30494
509         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
510         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
512         PR middle-end/30421
513         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
514         If lower_lastprivate_clauses emits some statements, append them
515         to dlist rather than body_p and to body_p append an initializer.
516         (lower_omp_for): Adjust caller.
518 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
520         * target.h (globalize_decl_name): New.
521         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
522         * output.h (default_globalize_decl_name): New.
523         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
524         globalize_label.
525         (globalize_decl): Ditto.
526         (default_globalize_decl_name): New.
527         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
528         (ia64_handle_version_id_attribute): New.
529         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
530         (ia64_asm_output_external): Use globalize_decl_name instead
531         of globalize_label.
532         * doc/extend.texi (version_id): New pragma.
533         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
535 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
537         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with 
538         _uleb128_t and _Unwind_SWord with _sleb128_t.
539         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
540         uw_frame_state_for, uw_update_context_1): Likewise.
541         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
542         * unwind-pe.h (read_uleb128, read_sleb128,
543         read_encoded_value_with_base): Likewise.
544         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
546 2007-01-24  Richard Guenther  <rguenther@suse.de>
548         * builtins.c (expand_builtin_cexpi): Get the fndecl
549         for cexp in the correct way.
551 2007-01-24  Jan Hubicka  <jh@suse.cz>
553         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
555 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
557         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
558         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
559         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
560         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
561         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
562         __SIZEOF_PTRDIFF_T__.
563         * doc/cpp.texi: Documentation for the new builtin macros added.
565 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
567         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
568         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
569         instead of operands[2] to avoid access past the end of array.
571 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
573         * reload1.c (emit_reload_insns): Pass the reload register
574         for a non-spill output reload through forget_old_reloads_1.
576 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
578         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
579         _SOFT_DOUBLE if doubles use software floating-point.
580         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
581         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
582         * config/rs6000/darwin-ldouble.c: Also compile functions for
583         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
584         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
585         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
586         commented-out long double override.
587         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
588         * config/rs6000/eabispe.h: Likewise.
589         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
590         long double for non-SPE.
591         (rs6000_handle_option): Likewise.
592         (invalid_e500_subreg): Disallow more subregs involding DImode,
593         DFmode, TImode or TFmode.
594         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
595         E500 double.
596         (legitimate_lo_sum_address_p): Also check for TFmode for E500
597         double.
598         (rs6000_legitimize_address): Also handle TFmode for E500 double.
599         (rs6000_legitimize_reload_address): Also handle TFmode for E500
600         double.
601         (rs6000_legitimate_address): Also check for TFmode for E500
602         double.
603         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
604         (spe_build_register_parallel): Handle TFmode and TCmode.
605         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
606         double.
607         (function_arg): Handle TFmode and TCmode for E500 double.
608         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
609         in general.
610         (print_operand): Handle TFmode and TImode for %y.
611         (rs6000_generate_compare): Handle TFmode comparisons for E500
612         double.
613         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
614         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
615         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
616         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
617         TFmode for E500 double.
618         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
619         (floatsidf2): Enable for E500 double.
620         (movtf_softfloat): Use rs6000_nonimmediate_operand.
621         (extenddftf2): Change to extenddftf2_fprs.
622         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
623         depending on TARGET_E500_DOUBLE.
624         (extendsftf2): Enable for E500 double.
625         (trunctfdf2): Enable for E500 double.
626         (trunctfsf2): Change to trunctfsf2_fprs.
627         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
628         depending on TARGET_E500_DOUBLE.
629         (floatsitf2): Enable for E500 double.
630         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
631         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
632         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
633         (negtf2): Change to negtf2_internal.
634         (negtf2): New expander.
635         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
636         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
637         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
638         (movdi_internal32): Use rs6000_nonimmediate_operand.
639         (unnamed splitter): Likewise.
640         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
641         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
642         (SPE64TF, DITI): New mode macros.
643         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
644         modes.
645         (frob_tf_ti): New.
646         (frob_<mode>_di_2): New.
647         (frob_tf_di_8_2): New.
648         (frob_di_df): Change to frob_di_<mode>; allow more modes.
649         (frob_ti_tf): New.
650         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
651         more modes.
652         (frob_ti_<mode>_8_2): New.
653         (frob_ti_tf_2): New.
654         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
655         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
656         TFmode.
657         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
658         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
659         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
660         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
661         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
662         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
663         tsttflt_gp): New.
665 2007-01-23  Ian Lance Taylor  <iant@google.com>
667         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
668         insn-*.o first.
669         (OBJS-archive): Reformat, alphabetize.
670         (OBJS): Change out_object_file to OBJS-md.
672 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
674         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
675         JUMP_P and CALL_P predicates where applicable.
676         * config/i386/i386.c: Ditto.
677         * config/i386/i386.md: Ditto.
678         * config/i386/mmx.md: Ditto.
679         * config/i386/predicates.md: Ditto.
681 2007-01-22  Andreas Schwab  <schwab@suse.de>
683         * config/m68k/m68k.h: Fix comment.
685 2007-01-22  Jan Hubicka  <jh@suse.cz>
687         * passes.c (init_optimization_passes): Do not rerun
688         pass_early_warn_uninitialized.
690 2007-01-22  Richard Guenther  <rguenther@suse.de>
692         PR tree-optimization/30038
693         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
694         function.
695         (execute_cse_sincos_1): Likewise.
696         (execute_cse_sincos): Likewise.
697         (gate_cse_sincos): Likewise.
698         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
699         the cexpi() canonicalization of sincos().
700         * tree-pass.h (pass_cse_sincos): Declare.
701         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
703 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
705         PR rtl-optimization/29329
706         * combine.c (replaced_rhs_insn): Rename to i2mod.
707         (replaced_rhs_value): Rename to i2mod_new_rhs.
708         (i2mod_old_rhs): New global variable.
709         (combine_instructions): Adjust for above change.  Save a copy of
710         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
711         note are substituted in the second instruction.
712         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
713         ditch the note if it pertains to the second eliminated register
714         and this register is mentioned in i2mod_old_rhs.
716         Revert:
717         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
719         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
720         the insn to which the note was originally attached.
722 2007-01-21  Jan Hubicka  <jh@suse.cz>
724         * ipa-inline.c (inlining_mode): Comment, move up.
725         (cgraph_decide_inlining_incrementally): Do not perform inlining
726         itself; fix handling of flattening of self recursive functions.
727         (cgraph_find_cycles): Remove.
728         (cgraph_flatten_node): Remove.
729         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
730         (try_inline): New function.
731         (cgraph_early_inlining): Update call of
732         cgraph_decide_inlining_incrementally.  Apply inlining here.
733         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
735 2007-01-21  Dirk Mueller  <dmueller@suse.de>
737         PR bootstrap/30511
738         * tree-vrp.c (check_array_bounds): do not warn
739         about ADDR_EXPR's of ARRAY_REF's which are immediately
740         used in binary expressions.
742 2007-01-21  Ira Rosen  <irar@il.ibm.com>
744         * tree-vectorizer.h (struct _stmt_vec_info): Add new field 
745         read_write_dep and macros for its access.
746         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
747         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove 
748         argument, call vect_check_interleaving for every independent pair of
749         data-refs. Mark loads that access the same memory location as a store
750         in the loop.
751         (vect_check_dependences): Remove.
752         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences 
753         call, fix the call to vect_analyze_data_ref_dependence.
754         (vect_analyze_data_ref_access): For statements that access the same 
755         data-ref, check that they are not stores; for loads, check that there
756         is no store that access the same location.
758 2007-01-20  Roger Sayle  <roger@eyesopen.com>
759             Joseph Myers  <joseph@codesourcery.com>
761         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
762         (-fdump-rtl-stack): Likewise.
763         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
764         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
765         (-mpt-fixed): Use @minus{} for minus sign.
766         (Using Precompiled Headers): Fix under/overfull hbox.
768 2007-01-20  Jan Hubicka  <jh@suse.cz>
770         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
771         place within annotation.
773 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
775         * pa.c (output_move_double): Change array size of xoperands to 4.
777 2007-01-20  Jan Hubicka  <jh@suse.cz>
779         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
780         as temporary.
782 2007-01-19  Ian Lance Taylor  <iant@google.com>
784         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
785         constant to unsigned HOST_WIDE_INT before negating.
787 2007-01-19  Ian Lance Taylor  <iant@google.com>
789         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
790         to unsigned.
791         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
792         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
793         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
794         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
795         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
796         (realloc_vdef, realloc_vuse): Change second parameter to
797         unsigned.
798         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
799         fields to unsigned.
800         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
801         unsigned. Change x and lim locals to unsigned.
802         (realloc_vdef, realloc_vuse): Change num_elem parameter to
803         unsigned.
804         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
805         (copy_virtual_operands): Change i and n locals to unsigned.
807 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
809         PR middle-end/29335
810         * builtins.c (fold_builtin_1): Handle builtin fdim.
812 2007-01-20  Jan Hubicka  <jh@suse.cz>
814         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
815         * tree-ssa-structalias.c (compute_points_to_sets): Do call
816         init_alias_heapvars.
817         (init_alias_heapvars): Initialize only when not already initialized.
818         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
820 2007-01-19  Roger Sayle  <roger@eyesopen.com>
822         * common.opt (fsigned-zeros): New command line option.
823         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
824         of flag_unsafe_math_optimizations.
825         * opts.c (set_fast_math_flags): The -ffast-math command line option
826         implies -fno-signed-zeros.
827         (fast_math_flags_set_p): Likewise.
829         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
830         the documentation of -ffast-math appropriately.  Wrap long lines.
832 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
834         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
835         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
837 2007-01-19  Tomas Bily  <tbily@suse.cz>
839         * cgraphunit.c (cgraph_finalize_function): Updating of pid
840         * tree-profile.c:
841         (tree_init_ic_make_global_vars): New function
842         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
843         (tree_gen_ic_profiler): New function
844         (tree_gen_ic_func_profiler): New function
845         (tree_profiling): Added calling of tree_gen_ic_func_profiler
846         (tree_profile_hooks): Added hook for indirec/virtual calls
847         * value-prof.c (tree_find_values_to_profile): New case for
848         indirect calls
849         (tree_values_to_profile): Call for determining indirect/virtual
850         counters
851         (tree_indirect_call_to_profile): New function
852         (tree_ic_transform): New function
853         (tree_ic): New function
854         (find_func_by_pid): New function
855         (init_pid_map): New function
856         (tree_value_profile_transformations): Added check for
857         indirect/virtual call transformation
858         * value-prof.h (enum hist_type): New counter type for
859         indirect/virtual calls
860         (profile_hooks): Added new hook for profiling indirect/virtual
861         calls
862         * profile.c (instrument_values): New case for indirect/virtual
863         call added
864         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
865         (GCOV_COUNTER_V_INDIR): New counter type
866         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
867         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
868         call added
869         * cgraph.c: Definition of cgraph_max_pid
870         (cgraph_create_node): Default init of pid attribute
871         * cgraph.h: Declaration of cgraph_max_pid
872         (struct cgraph_node): Added pid attribute
873         * libgcov.c (__gcov_indirect_call_profiler): New function
874         (__gcov_one_value_profiler_body): New function
875         (__gcov_one_value_profiler): Body was moved to
876         __gcov_one_value_profiler_body and calls it
878 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
879         
880         * doc/gty.texi (Options): Document the mark_hook option to GTY.
881         * gengtype.c (write_types_data, write_func_for_structure,
882         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
883         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
884         (walk_type, write_func_for_structure): Generate the mark_hook if
885         needed.
887 2007-01-19  Jan Hubicka  <jh@suse.cz>
889         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
890         'early' argument take inlining mode argument specifying whether to
891         inline for size/speeed or all functions; add support for flattening;
892         improve dumpting.
893         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
895 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
897         PR c++/17947
898         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
899         the name of the declared identifier.
900                 
901 2007-01-19  Dirk Mueller  <dmueller@suse.de>
903         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
904         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
905         duplicate evaluation.
907 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
909         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
910         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
911         to XFMode, use acosxf2 and truncate result to requested mode.
912         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
913         (asin<mode>2): Similarly, with asin expanders.
914         (*fscalexf4_i387): Rename from *fscalexf4.
915         (expNcorexf3): New expander.
916         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
917         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
918         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
919         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
920         patterns for SSE math.
921         (exp10<mode>2): Similarly, with exp10 expanders.
922         (exp2<mode>2): Similarly, with exp2 expanders.
923         (expm1<mode>2): Similarly, with expm1 expanders.
924         (ldexp<mode>3): Similarly, with ldexp expanders.
925         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
926         Use gen_truncxf<mode>2_i387_noop to truncate result.
928 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
930         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
931         for ColdFire targets.
933 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
934             Richard Sandiford  <richard@codesourcery.com>
936         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
937         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
938         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
939         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
940         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
941         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
943 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
945         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
946         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
947         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
948         (*adddi_dilshr32_cf): New pattern.
949         (adddi3, subdi3): Remove first alternatives.
951 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
953         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
954         implemented using move.l, do not use its cc result for floating-point
955         comparisons.
957 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
959         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
960         from accepting 's' constraints if flag_pic.
962 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
964         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
965         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
966         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
967         condition from TARGET_68881 to TARGET_HARD_FLOAT.
969 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
971         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
972         ColdFire alternatives.
973         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
974         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
975         architectures that support ff1 instruction.
977 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
978             Julian Brown  <julian@codesourcery.com>
980         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
981         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
982         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
983         Add commentary.
985 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
987         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
988         and return a bool.
989         (output_move_const_into_data_reg, output_move_simode_const): Delete.
990         * config/m68k/m68k.c (const_method, const_int_cost): Take a
991         HOST_WIDE_INT instead of an rtx.
992         (m68k_rtx_costs): Update call accordingly.
993         (output_move_const_into_data_reg): Likewise.  Fix formatting.
994         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
995         Return a bool.
996         (output_move_simode_const): Update calls after above changes.
997         Rework to use automatic variables and predicates like MEM_P.
998         * config/m68k/m68k.md (pushexthisi_const): Update call to
999         valid_mov3q_const.
1001 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1003         * tree-ssa-alias.c (perform_var_substitution): Fix typo
1004         in dump_flags test.
1006 2007-01-19  Richard Guenther  <rguenther@suse.de>
1008         * builtins.c (expand_builtin_cexpi): Fall back to expanding
1009         via cexp in case sincos is not available.
1011 2007-01-19  Richard Guenther  <rguenther@suse.de>
1013         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1014         * defaults.h (TARGET_HAS_SINCOS): Default to off.
1015         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1016         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1017         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1018         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1019         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1020         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1022 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1024         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1025         New insn patterns.
1026         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1027         (atan2xf3): Directly generate RTL pattern.
1028         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1029         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1030         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1031         disable patterns for SSE math.
1032         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1033         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1034         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1035         disable patterns for SSE math.
1037 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1039         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1041 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1043         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1044         VIEW_CONVERT_EXPR to the same type as it's operand.
1046 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1048         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1049         configured for long double 128.
1051 2007-01-18  Mike Stump  <mrs@apple.com>
1053         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1054         error.
1056 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1058         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1059         type correct.
1061 2007-01-18  Jan Hubicka  <jh@suse.cz>
1063         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1064         greater than NUM_VOP_FREE_BUCKETS.
1066 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1068         * tree-ssa-structalias.c: Update comments.
1069         (ptabitmap_obstack): Removed.
1070         (pta_obstack): New.
1071         (oldpta_obstack): Ditto.
1072         (stats): Add a few members.
1073         (struct variable_info): Remove node, complex, address_taken, and
1074         indirect_target members. Add oldsolution member.
1075         (new_var_info): Do not initialize removed members.
1076         (constraint_expr_type): Remove INCLUDES.
1077         (constraint_graph): Add size, implicit_preds, rep,
1078         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1079         (FIRST_REF_NODE): New macro.
1080         (LAST_REF_NODE): Ditto.
1081         (FIRST_ADDR_NODE): Ditto.
1082         (find): New function.
1083         (unite): Ditto.
1084         (dump_constraint): Do not handle INCLUDES.
1085         (insert_into_complex): Do not insert duplicate constraints.
1086         (condense_varmap_nodes): Renamed and rewritten into ...
1087         (merge_node_constraints): This. Also fix bug in handling of
1088         offseted copy constraints.
1089         (clear_edges_for_node): No longer need to deal with preds at all,
1090         or removing associated preds/succs. 
1091         (merge_graph_nodes): Deal with indirect_cycles.
1092         Don't deal with predecessors.
1093         (add_implicit_graph_edge): New function.
1094         (add_pred_graph_edge): Ditto.
1095         (add_graph_edge): Don't deal with predecessors.
1096         (build_constraint_graph): Removed.
1097         (build_pred_graph): New function.
1098         (build_succ_graph): Ditto.
1099         (struct scc_info): Removed in_component. Added roots, dfs, and
1100         node_mapping. Remove visited_index, unification_queue.
1101         (scc_visit): Deal with union-find we do now.
1102         Deal with cycles with REF nodes.
1103         (collapse_nodes): Renamed and rewritten to ...
1104         (unify_nodes): This.
1105         (process_unification_queue): Removed.
1106         (topo_visit): Cleanup
1107         (do_da_constraint): Use find.
1108         (do_sd_constraint): Ditto.
1109         (do_ds_constraint): Ditto.
1110         (do_complex_constraint): Ditto.
1111         (init_scc_info): Update for removed and added members.
1112         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1113         (find_indirect_cycles): This.
1114         (equivalence_class): New variable.
1115         (label_visit): New function.
1116         (perform_variable_substitution): Rewritten.
1117         (free_var_substitution_info): New function.
1118         (find_equivalent_node): Ditto.
1119         (move_complex_constraints): Ditto.
1120         (eliminate_indirect_cycles): Ditto.
1121         (solve_graph): Only propagate changed bits.
1122         Use indirect cycle elimination.
1123         Use find.
1124         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1125         (tree_id_eq): Renamed to ...
1126         (tree_vi_eq): This. Update for member change
1127         (insert_id_for_tree): Renamed and rewritten to ...
1128         (insert_vi_for_tree): This.
1129         (lookup_id_for_tree): Renamed and rewritten to ...
1130         (lookup_vi_for_tree): This.
1131         (get_id_for_tree): Renamed and rewritten to ...
1132         (get_vi_for_tree): Ditto.
1133         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1134         (process_constraint): Don't handle INCLUDES.
1135         Remove special ADDRESSOF case.
1136         (find_func_aliases): Rewrite to use vi functions instead of id
1137         ones.
1138         (create_function_info_for): Ditto.
1139         (create_variable_info_for): Ditto.
1140         (intra_create_variable_infos): Ditto.
1141         (merge_smts_into): Ditto.
1142         (find_what_p_points_to): Ditto.
1143         (init_base_vars): Ditto.
1144         (init_alias_vars): Ditto.
1145         (remove_preds_and_fake_succs): New function.
1146         (dump_sa_points_to_info): Dump new stats.
1147         (dump_solution_for_var): Use find.
1148         (set_used_smts): Fix formatting.
1149         (compute_points_to_sets): Updated for new functions.
1150         (ipa_pta_execute): Ditto.
1152 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1153             Richard Sandiford  <richard@codesourcery.com>
1155         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1156         parallels.
1157         * calls.c (expand_call): If the return value is a PARALLEL,
1158         extract its first member.
1159         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1160         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1161         use FUNCTION_EXTRA_EPILOGUE.
1162         (m68k_function_value): Return a PARALLEL if the return value
1163         is of a pointer type.
1164         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1165         (FUNCTION_EXTRA_EPILOGUE): Remove.
1166         * config/m68k/m68k.md (D0_REG): New constant.
1168 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1170         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1171         output a NOP for empty epilogues.
1173 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1175         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1176         before function.  Extend register save check to include all
1177         registers, not just integer ones.
1179 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1181         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1182         (m68k_use_return_insn): ...this.
1183         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1184         * config/m68k/m68k.c (use_return_insn): Rename to...
1185         (m68k_use_return_insn): ...this.
1186         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1187         of USE_RETURN_INSN.
1189 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1191         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1192         nearest even, fix denormal rounding overflow.
1194 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1196         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1197         f%$smove and f%$move.
1198         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1199         (extendsfdf2_cf): Use fdmove instead of f%&move.
1200         (truncdfsf2_cf): Use fsmove instead of f%$smove.
1201         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1203 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1205         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1206         GPR<-GPR moves.
1208 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1210         * real.h (real_format): Add a canonical_nan_lsbs_set field.
1211         (coldfire_single_format): Declare.
1212         (coldfire_double_format): Likewise.
1213         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1214         of qnan_msb_set to determine the lower bits of a canonical
1215         NaN significand.
1216         (encode_ieee_double): Likewise.
1217         (encode_ieee_quad): Likewise.
1218         (ieee_single_format): Initialize canonical_nan_lsbs_set.
1219         (mips_single_format): Likewise.
1220         (ieee_double_format): Likewise.
1221         (mips_double_format): Likewise.
1222         (ieee_extended_motorola_format): Likewise.
1223         (ieee_extended_intel_96_format): Likewise.
1224         (ieee_extended_intel_128_format): Likewise.
1225         (ieee_extended_intel_96_round_53_format): Likewise.
1226         (ibm_extended_format): Likewise.
1227         (mips_extended_format): Likewise.
1228         (ieee_quad_format): Likewise.
1229         (mips_quad_format): Likewise.
1230         (vax_f_format): Likewise.
1231         (vax_d_format): Likewise.
1232         (vax_g_format): Likewise.
1233         (i370_single_format): Likewise.
1234         (i370_double_format): Likewise.
1235         (decimal_single_format): Likewise.
1236         (decimal_double_format): Likewise.
1237         (decimal_quad_format): Likewise.
1238         (c4x_single_format): Likewise.
1239         (c4x_extended_format): Likewise.
1240         (real_internal_format): Likewise.
1241         (coldfire_single_format): New real_format.
1242         (coldfire_double_format): Likewise.
1243         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1244         canonical_nan_lsbs_set.
1245         (pdp11_d_format): Likewise.
1246         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1247         if TARGET_COLDFIRE_CPU.
1249 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1251         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1252         (output_call, m68k_legitimize_call_address): Declare.
1253         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1254         parenthesees.  Add support for a 'W' constraint.
1255         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1256         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1257         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1258         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1259         variables.
1260         (override_options): Initialize them.  Do not set flag_no_function_cse
1261         for TARGET_ID_SHARED_LIBRARY.
1262         (m68k_output_pic_call): Delete.
1263         (m68k_legitimize_call_address): New function.
1264         (print_operand): Remove the %o prefix.  Handle the %p prefix.
1265         (output_call): New function.
1266         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
1267         target address from the GOT if symbolic jumps are not allowed.
1268         * config/m68k/m68k.md (call, general_operand): Do not set
1269         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
1270         Merge separate flag_pic and !flag_pic define_insns into...
1271         (*call, *call_value): ...these new patterns.  Match the address
1272         rather than the containing MEM and require it to be a call_operand.
1273         Use output_call to generate the asm template.
1274         * config/m68k/predicates.md (const_call_operand): New predicate.
1275         (call_operand): Likewise.
1277 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1279         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1280         (STACK_GROWS_DOWNWARD): Define to 1.
1281         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1282         comments, emphasizing that these values are only defaults.
1283         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1284         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1286 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1288         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1290 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1291             Richard Sandiford  <richard@codesourcery.com>
1293         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1294         (FP_REGNO_P): New macros.
1295         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1296         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1297         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1298         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1300 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1302         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1303         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1304         to $tmake_file.
1305         * config/m68k/t-floatlib: New file, extracting common code from...
1306         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1307         * config/m68k/t-uclinux: Here.
1308         * config/m68k/fpgnulib.c: Do not compile extendeed precision
1309         routines on ColdFire targets.
1311 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1313         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1314         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1315         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1316         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1317         explicitly set MOTOROLA to 1.
1318         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1320 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1321             Nathan Sidwell  <nathan@codesourcery.com>
1323         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1324         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1325         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1326         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1327         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1328         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1329         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1330         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1331         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1332         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1333         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1334         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1335         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1336         (DBX_REGISTER_NUMBER): Delete.
1337         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1338         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1339         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1341 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1343         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1344         * config/flat.h: New file.
1345         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1346         OBJECT_FORMAT_FLAT.
1347         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1348         indirect references for -msep-data or -mid-shared-library.
1349         Do not use PC-relative code addresses either.
1350         * config/m68k/m68k.c (override_options): Restrict -fPIC error
1351         to -mpcrel.
1352         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
1353         for shared libraries and crt1.o for executables.  Use crti.o and
1354         crtbegin.o.
1355         (ENDFILE_SPEC): Use crtend.o and crtn.o.
1356         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1357         Do not add -elf2flt or -shared-lib-id options here.
1358         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
1359         Pass -shared-lib-id if -mid-shared-library, taking the library
1360         identifier from -mshared-library-id if given, otherwise
1361         defaulting to 0.
1362         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1363         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1364         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1365         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1366         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1367         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1368         and crtend.o.
1369         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1370         jump sequence for ISA A and ISA A+.
1371         (PICJUMP): Likewise.
1373 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1375         * config.gcc (m68k-*-uclinux*): Base the port on the common
1376         and m68k GNU/Linux files rather than on the generic ELF ones.
1377         * config/m68k/uclinux.h (TARGET_VERSION): Override.
1378         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1380 2007-01-18  Julian Brown  <julian@codesourcery.com>
1381             Richard Sandiford  <richard@codesourcery.com>
1383         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1384         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1385         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1386         before redefining.
1387         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1388         80 unconditionally.
1389         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1391 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1393         * doc/install.texi: Document m68k-uclinuxoldabi.
1394         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1395         * config/m68k/uclinux-oldabi.h: New file, copied from
1396         config/m68k/uclinux.h.
1398 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
1400         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1401         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1402         (m32r_compute_frame_size): Use unsigned for regno.
1403         (m32r_reload_lr): Use gen_frame_mem.
1404         (pop): New.
1405         (m32r_output_function_epilogue): Don't output the function epilogue
1406         textually here.
1407         (m32r_expand_epilogue): New.
1408         (direct_return): Return false if the function has the interrupt
1409         attribute.
1410         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1411         * config/m32r/m32r.md (epilogue): New expander.
1412         (return_lr, return_rte): New insns.
1413         (return): Make it expander.
1414         (return_normal): New expander.
1416 2007-01-18  Josh Conner  <jconner@apple.com>
1418         PR target/30485
1419         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1420         support for UNLE, UNLT, UNGE, and UNGT.
1422 2007-01-18  Jan Hubicka  <jh@suse.cz>
1424         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1425         * tree-tailcall (add_virtual_phis): Likewise.
1426         (optimize_tail_call): Return TODOs.
1427         (execute_tail_calls): Return TODOs.
1428         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1429         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1430         changed.
1431         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1432         Cleanup cfg using TODO.
1433         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1434         * passes.c (execute_function_todo): When cleanup did something, remove
1435         unused locals.
1436         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1437         (make_edges): Don't cleanup_cfg.
1439 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
1441         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1442         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1443         operate on the top of stack.
1444         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1445         register handling of instructions that output to the second
1446         stack slot.
1447         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1448         (move_for_stack_reg): Special-case check for dead destination
1449         stack slot for constant load of 1.0 inside UNSPEC_TAN.
1451         * config/i386/i386.md (UNSPEC_TAN): New constant.
1452         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1453         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1454         to correctly model move of constant 1.0 to top stack slot.
1455         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1456         (unnamed peephole2 pattern): Remove corresponding peephole2
1457         pattern that optimizes tan insn and loading of constant 1.0.
1458         (tanxf2): Use fptanxf4_i387.
1459         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1460         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
1461         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1462         disable patterns for SSE math.
1463         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1464         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1466 2007-01-18  Dirk Mueller  <dmueller@suse.de>
1467             Richard Guenther <rguenther@suse.de>
1469         PR diagnostic/8268
1470         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1471         * common.opt (Warray-bounds): Add new warning option.
1472         * c-opts.c (c_common_handle_option): Define -Warray-bounds
1473         if -Wall is given.
1474         * Makefile.in: make tree-vrp.o depend on toplev.h
1475         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1476         is enabled.
1477         (check_array_refs, check_array_bounds, check_array_ref): New.
1479 2007-01-18  Jan Hubicka  <jh@suse.cz>
1481         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1482         (execute_ssa_ccp): Return flags conditionally.
1483         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1484         changed.
1485         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1487 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
1489         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1490         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1492         * Makefile.c (GTFILES): Add cfglayout.h.
1493         * gengtype.c (open_base_files): Likewise.
1494         * cfglayout.c (cfg_layout_function_footer,
1495         cfg_layout_function_header) Reindent to make gengtype happy.
1496         * cfglayout.h (cfg_layout_function_footer,
1497         cfg_layout_function_header): Add GTY(()) marker.
1499         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1501 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1503         * genautomata.c (write_automata): Include xstrerror output in the
1504         error message if writing the DFA description file fails.
1506 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1508         * config/mips/mips-protos.h (mips_output_external): Make it
1509         return void.
1510         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1511         * config/mips/mips.c (irix_output_external_libcall): Likewise.
1512         (extern_list): Likewise.
1513         (extern_head): Likewise.
1514         (TARGET_ASM_FILE_END): Likewise.
1515         (mips_file_end): Likewise.
1516         (mips_output_external): Rewritten.
1518 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1520         * genpreds.c (write_insn_preds_c): Only write out the function
1521         body for regclass_for_constraint if we have register constraints.
1523 2007-01-17  Tom Tromey  <tromey@redhat.com>
1525         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1526         * doc/install.texi (Testing): Use sourceware.org.
1527         (Binaries): Likewise.
1528         (Specific): Likewise.
1529         * doc/contrib.texi (Contributors): Use sourceware.org.
1531 2007-01-17  Anatoly Sokolov <aesok@post.ru>
1533         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1534         (AVR_ENHANCED): Rename to ...
1535         (AVR_HAVE_MUL): ... new.
1536         (avr_enhanced_p): Rename to ...
1537         (avr_have_mul_p): ... new.
1538         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of 
1539         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1540         * config/avr/avr.c (avr_enhanced_p): Rename to ...
1541         (avr_have_mul_p): ... new.
1542         (base_arch_s): Rename 'enhanced' to 'have_mul'.
1543         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1544         'avr_enhanced_p' and 'enhanced'.
1545         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use 
1546         AVR_HAVE_MUL instead of AVR_ENHANCED.
1547         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1548         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3, 
1549         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL 
1550         instead of AVR_ENHANCED.
1551         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1552         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1553         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1555 2007-01-17  Ian Lance Taylor  <iant@google.com>
1557         * vec.h (VEC_reserve_exact): Define.
1558         (vec_gc_p_reserve_exact): Declare.
1559         (vec_gc_o_reserve_exact): Declare.
1560         (vec_heap_p_reserve_exact): Declare.
1561         (vec_heap_o_reserve_exact): Declare.
1562         (VEC_OP (T,A,reserve_exact)): New static inline function, three
1563         versions.
1564         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1565         negative parameter.
1566         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1567         (VEC_OP (T,A,copy)) [all versions]: Likewise.
1568         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1569         * vec.c (calculate_allocation): Add exact parameter.  Change all
1570         callers.
1571         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1572         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1573         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1574         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1575         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1576         (vec_heap_p_reserve_exact): New function.
1577         (vec_heap_o_reserve_exact): New function.
1579 2007-01-17  Jan Hubicka  <jh@suse.cz>
1581         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1583 2007-01-17  Eric Christopher  <echristo@apple.com>
1585         * config.gcc: Support core2 processor.
1587 2007-01-16  Jan Hubicka  <jh@suse.cz>
1589         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1590         the names, just unlink the chain so we don't crash on dangling pointers
1591         to dead SSA names.
1593 2007-01-16  Jan Hubicka  <jh@suse.cz>
1595         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1596         * tree-pass.h: Reorder to make IPA passes appear toegher.
1597         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1598         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1599         parameters, do not call early inliner.
1600         * ipa-inline.c: Update comments.  Include tree-flow.h
1601         (cgraph_decide_inlining): Do not compute inlining parameters.
1602         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1603         be called with function context set up.
1604         (pass_ipa_inline): Remove unreachable functions before pass.
1605         (cgraph_early_inlining): Simplify assuming to be called from the
1606         PM as local pass.
1607         (pass_early_inline): New pass.
1608         (cgraph_gate_ipa_early_inlining): New gate.
1609         (pass_ipa_early_inline): Turn into simple wrapper.
1610         (compute_inline_parameters): New function.
1611         (gate_inline_passes): New gate.
1612         (pass_inline_parameters): New pass.
1613         (apply_inline): Move here from tree-optimize.c
1614         (pass_apply_inline): New pass.
1615         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1616         transforming.
1617         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1618         doing them by hand.
1619         (tree_function_versioning): Do not allocate dummy struct function.
1620         * tree-inline.h (optimize_inline_calls): Update prototype.
1621         * tree-optimize.c (execute_fixup_cfg): Export.
1622         (pass_fixup_cfg): Remove
1623         (tree_rest_of_compilation): Do not apply inlines.
1624         * tree-flow.h (execute_fixup_cfg): Declare.
1625         * Makefile.in (gt-passes.c): New.
1626         * passes.c: Include gt-passes.h
1627         (init_optimization_passes): New passes.
1628         (nnodes, order): New static vars.
1629         (do_per_function_toporder): New function.
1630         (execute_one_pass): Dump current pass here.
1631         (execute_ipa_pass_list): Don't dump current pass here.
1633 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
1635         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1636         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1638 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
1640         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1641         (strong_alias): Define.
1642         (__gcc_qmul): Provide non-FMA for soft-float.
1643         (__gcc_qdiv): Same.
1644         (__gcc_qneg): New.
1645         (__gcc_qeq): New.
1646         (__gcc_qle): New.
1647         (__gcc_qge): New.
1648         (__gcc_qunord): New.
1649         (__gcc_stoq): New.
1650         (__gcc_dtoq): New.
1651         (__gcc_qtos): New.
1652         (__gcc_qtod): New.
1653         (__gcc_qtoi): New.
1654         (__gcc_qtou): New.
1655         (__gcc_itoq): New.
1656         (__gcc_utoq): New.
1657         (fmsub): New.
1658         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1659         soft-float functions.
1660         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1661         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1662         about long double soft float.
1664 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
1665             Tehila Meyzels  <tehila@il.ibm.com>
1667         * tree-vectorizer.h (is_pattern_stmt_p): New.
1668         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1669         formatting (tabs instead of spaces). Cleanup and clarify setting
1670         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1671         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1672         (vectorizable_type_demotion): Check that types are integral.
1673         (vectorizable_type_promotion): Likewise.
1674         (vectorizable_store): Fix typo.  Eliminate new-line at end of
1675         comments.
1677 2007-01-16  Jan Hubicka  <jh@suse.cz>
1679         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1680         cgraph edges too.
1682 2007-01-15  Eric Christopher  <echristo@apple.com>
1684         * ifcvt.c: Include vec.h, vecprim.h.
1685         (check_cond_move_block): New argument regs.
1686         Reorganize. Add registers used to regs.
1687         (cond_move_process_if_block): Use regs set above as
1688         loop bounds.
1690 2007-01-15  Eric Christopher  <echristo@apple.com>
1692         * config/darwin.h: Update copyright.
1693         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1694         (LINK_COMMAND_SPEC): Add -u.
1695         (LINK_SPEC): Fix umbrella for above.
1697 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
1699         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1700         glibc CVS.
1702 2007-01-15  Tom Tromey  <tromey@redhat.com>
1704         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1705         * doc/install.texi (Testing): Don't mention jacks.
1706         (Configuration): Document --enable-java-maintainer-mode.  Move
1707         --with-java-home to libgcj-specific section.  Document
1708         --with-ecj-jar.
1709         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1711 2007-01-15  Jan Hubicka  <jh@suse.cz>
1713         * tree-ssa-dce.c (DCE_TODOs): New.
1714         (propagate_necessity): Return if something changed.
1715         (eliminate_unnecessary_stmts): Likewise.
1716         (perform_tree_ssa_dce): Return TODO flags when needed.
1717         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1719 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
1721         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1722         (fyl2x_extend<mode>xf3_i387): New insn pattern.
1723         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1724         patterns using X87MODEF12 mode macro.  Extend operand 1
1725         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1726         (log10<mode>2): Ditto.
1727         (log2<mode>2): Ditto.
1728         (log1p<mode>2): Ditto.
1729         (logb<mode>2): Ditto.
1730         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1731         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1732         (*fxtractxf3_i387): Rename from *fxtractxf3.
1733         (fxtract_extend<mode>xf3_i387): New insn pattern.
1734         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1735         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1736         and gen_fyl2xxf3_i387().
1738 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
1740         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1741         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1742         get_loop_exit_edges.  Changed return type to bool.
1743         * cfgloop.h (fix_loop_placement): Declaration removed.
1745 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1747         * param.h (MIN_VECT_LOOP_BOUND): New.
1748         * params.def (MIN_VECT_LOOP_BOUND): New.
1749         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1750         argument - minimum threshold for number of iterations.
1751         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1752         argument to declaration.
1753         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1754         MIN_VECT_LOOP_BOUND.
1755         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1756         slpeel_tree_peel_loop_to_edge with additional argument.
1757         (vect_do_peeling_for_alignment): Likewise.
1758         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1760 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1762         PR target/30413
1763         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1764         operands of size 1.
1766 2007-01-14  Jan Hubicka  <jh@suse.cz>
1768         * tree-dfa.c (remove_referenced_var): New function.
1769         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1770         prune referenced vars list too.
1771         * tree-flow.h (remove_referenced_var): Declare.
1773 2007-01-14  Jan Hubicka  <jh@suse.cz>
1775         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1776         separately.
1777         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1778         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1779         * except.h (verify_eh_throw_table_statements): Kill prototype.
1780         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1781         into call exrepssions.
1782         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1783         eh trhow verifier.
1784         * tree-cfg.c: Include pointer-set.h.
1785         (verify_node_sharing): Work on pointer set.
1786         (verify_eh_throw_stmt_node): New.
1787         (verify_stmts): Use pointers sets, verify throw_stmt.
1789 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1791         * ipa-reference.c (analyze_function): Consider also addresses taken
1792         in phi nodes.
1794 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1796         * c-typeck.c (null_pointer_constant_p): Replace use of
1797         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1798         (build_c_cast): Likewise.
1800 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1802         * tree.h (force_fit_type_double): Remove unused final argument.
1803         * c-common.c (constant_expression_warning): Replace use of
1804         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1805         (convert_and_check): Likewise.
1806         (shorten_compare): Update call to force_fit_type_double.
1807         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1808         * convert.c (convert_to_pointer): Update call to
1809         force_fit_type_double.
1810         * fold-const.c (force_fit_type_double): Remove overflowed_const
1811         argument.
1812         (int_const_binop, fold_convert_const_int_from_int,
1813         fold_convert_const_int_from_real, fold_div_compare,
1814         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1815         fold_abs_const, fold_not_const): Remove the final argument from
1816         calls to force_fit_type_double.
1818 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1820         * configure.ac: Set insn to "nop" for spu-*-* also.
1821         * configure: Regenerate.
1823 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
1825         * builtins.def : Add BUILT_IN_FREE.
1827 2007-01-12  Jan Hubicka  <jh@suse.cz>
1829         PR tree-optimization/30443
1830         * tree-inline.c (tree_function_versioning): Do not optimize when
1831         cloning for inlining.
1833 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
1835         * doc/loop.texi: Document recording of loop exits.
1836         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1837         (update_single_exits_after_duplication,
1838         update_single_exit_for_duplicated_loop,
1839         update_single_exit_for_duplicated_loops): Removed.
1840         (duplicate_loop_to_header_edge): Do not call
1841         update_single_exits_after_duplication and
1842         update_single_exit_for_duplicated_loops.
1843         (loop_version): Do not update single_exit information.
1844         (fix_loop_structure): Use record_loop_exits instead of
1845         mark_single_exit_loops.
1846         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1847         the lists of loop exits.
1848         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1849         split_edge, merge_blocks): Update the lists of loop exits.
1850         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1851         loop_optimizer_init.
1852         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1853         of mark_single_exit_loops.
1854         (loop_optimizer_finalize): Call release_recorded_exits.
1855         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1856         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1857         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1858         update single exit information.
1859         * lambda-code.c (perfect_nestify): Ditto.
1860         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1861         (mark_single_exit_loops): Removed.
1862         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1863         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1864         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1865         functions.
1866         (get_loop_exit_edges, single_exit): Use recorded exit lists.
1867         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1868         (verify_loop_structure): Verify consistency of the exit lists.
1869         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
1870         (set_single_exit): Removed.
1871         * cfgloop.h (struct loop_exit): New function.
1872         (struct loop): single_exit_ field replaced by exits field.
1873         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1874         (struct loops): Added exits hash.
1875         (mark_single_exit_loops, set_single_exit): Declaration removed.
1876         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1878 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1880         * doc/invoke.texi: Avoid use of @headitem.
1882 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1884         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1886 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1888         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1889         list of targets.
1891 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1892             Richard Sandiford  <richard@codesourcery.com>
1894         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1895         * config/m68k/m68k.opt: Resort options.
1896         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1898 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1899             Richard Sandiford  <richard@codesourcery.com>
1901         * doc/invoke.texi: Document the macros that are defined by
1902         m68k's -mtune and -mhard-float options.
1903         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1904         (m68k_cpp_cpu_family): Likewise.
1905         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1906         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
1907         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
1908         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1909         (m68k_cpp_cpu_family): Likewise.
1911 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1913         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1914         macros besides mc68000 as tuning macros.  Use a switch statement
1915         to set them and mcpu32.
1917 2007-01-12  Julian Brown  <julian@codesourcery.com>
1919         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1920         (TARGET_68040_ONLY): Rename to...
1921         (TARGET_68040): ...this.
1922         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1923         * config/m68k/m68k.md: Likewise.
1925 2007-01-12  Julian Brown  <julian@codesourcery.com>
1926             Nathan Sidwell  <nathan@codesourcery.com>
1927             Richard Sandiford  <richard@codesourcery.com>
1929         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1930         the -mcpu= argument associated with the --with-cpu setting.
1931         Define M68K_DEFAULT_TUNE to the default -mtune= option,
1932         if different from the one implied by the -mcpu setting.
1933         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1934         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
1935         * doc/invoke.texi: Mention ColdFire in the introduction to the
1936         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
1937         -mno-div and -mhard-float options.  Make -m68881 a synonym for
1938         -mhard-float.  Document the previously-undocumented -m5206e,
1939         -m528x, -m5307 and -m5407 options.  Tweak the existing option
1940         documentation for consistency.
1941         * doc/install.texi: Mention new --with-cpu arguments.
1942         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1943         default CPU if neither -mcpu nor -march are specified.
1944         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1945         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1946         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
1947         using m68k_tune to decide between families that implement the
1948         same ISA.  Use m68k_tune to set __mcfv4e__.
1949         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1950         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1951         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1952         (FL_ISA_C, FL_ISA_MMU): New macros.
1953         (MASK_COLDFIRE): Delete.
1954         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1955         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1956         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1957         (TARGET_HARD_FLOAT): Do not define here.
1958         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1959         (TUNE_68000): New macro.
1960         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1961         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1962         (TUNE_CFV2): Redefine in terms of m68k_tune.
1963         (uarch_type, target_device, fpu_type): New enums.
1964         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1965         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1966         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1967         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1968         (FL_FOR_isa_c): New macros.
1969         (m68k_isa): New enum.
1970         (m68k_target_selection): New structure.
1971         (all_devices, all_isas, all_microarchs): New tables.
1972         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1973         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1974         (MASK_ALL_CPU_BITS): Delete.
1975         (m68k_find_selection): New function.
1976         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1977         Map the legacy target options to a combination of the new ones.
1978         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1979         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
1980         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1981         settings.
1982         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1983         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1984         properties.
1985         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1986         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1987         * config/m68k/m68k-devices.def: New file.
1989 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1990             Nathan Sidwell  <nathan@codesourcery.com>
1992         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1993         (SUBTARGET_EXTRA_SPECS): New macros.
1994         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1995         use %(asm_cpu_spec) instead.
1996         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1997         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1998         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1999         (EXTRA_SPECS): Rename to...
2000         (SUBTARGET_EXTRA_SPECS): ...this.
2002 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2003             Richard Sandiford  <richard@codesourcery.com>
2004             Julian Brown  <julian@codesourcery.com>
2006         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2007         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2008         to the configuration's default CPU.
2009         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2010         Remove default masks.
2011         (m680[012]0-*-*): Set the default with_cpu to the first part of
2012         the target name.
2013         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2014         (m68k*-*-linux): Extend the --with-cpu handling to...
2015         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
2016         and m68010.  Don't set target_cpu_default2.
2017         * doc/install.texi: Document --with-cpu for m68k.
2018         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2019         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2020         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2021         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2022         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2023         (ASM_SPEC): Remove use of %(asm_cpu_default).
2024         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2025         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2026         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2027         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2028         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2029         (TARGET_DEFAULT): Delete.
2030         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2031         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2032         (CPP_CPU_SPEC): Delete.
2033         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2034         the appropriate string.
2035         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2036         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2037         (ASM_SPEC): Don't use %(asm_default_spec).
2038         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2039         TARGET_DEFAULT and add MASK_68881.
2040         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2042 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2044         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2045         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2046         MASK_68010 alongside MASK_68020.
2047         * doc/invoke.texi: Document -m68010.
2048         * config/m68k/m68k.opt (m68010): New.
2049         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2050         if TUNE_68010.
2051         (TUNE_68010): New macro.
2052         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2053         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2054         (M68K_CPU_m68332): Add MASK_68010.
2055         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2056         fallback definition.
2057         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2058         defines.
2059         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2060         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2061         to all entries that use MASK_68020.
2062         (output_move_simode_const, output_move_himode, output_move_qimode)
2063         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2064         instead of TARGET_68020 to select clr behavior.  Remove comment
2065         about there being no TARGET_68010.
2066         * config/m68k/m68k.md: Likewise throughout.
2068 2007-01-12  Julian Brown  <julian@codesourcery.com>
2070         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2071         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2072         * config/m68k/m68k.md: Likewise.
2074 2007-01-12  Julian Brown  <julian@codesourcery.com>
2076         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2077         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2079 2007-01-12  Julian Brown  <julian@codesourcery.com>
2081         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2082         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2083         * config/m68k/m68k.md: Likewise.
2085 2007-01-12  Julian Brown  <julian@codesourcery.com>
2086             Richard Sandiford  <richard@codesourcery.com>
2088         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2089         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2090         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2091         TARGET_CPU32.
2092         (TARGET_CPU32): Rename to...
2093         (TUNE_CPU32): ...this.
2094         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2095         (TUNE_CFV2): New macros.
2096         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2097         remove conditions that are implied by TARGET_68020.
2098         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2099         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2100         (m68k_output_function_epilogue): Likewise.
2101         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2102         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2103         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2104         non-68000 timings.  Refactor multiplication and division costs.
2105         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2106         TUNE_CPU32 instead of TARGET_CPU32.
2107         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2108         and TUNE_68060 instead of TARGET_68060.
2109         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2110         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2111         TARGET_CPU32.
2112         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2113         && !TARGET_COLDFIRE" to choose between moveq and clr.
2114         Likewise in the unnamed movsf pattern.
2115         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2116         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2117         ashiftrt pattern.
2119 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2121         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2122         of tabbing before backslashes.
2124 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2126         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2127         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2128         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2130 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2132         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2133         find_interesting_uses_cond.
2134         (find_interesting_uses_cond): Use extract_cond_operands.
2135         (rewrite_use_compare): Use extract_cond_operands and
2136         force_gimple_operand_bsi.  Do not call update_stmt.
2137         (determine_use_iv_cost_condition): Use extract_cond_operands.
2138         Return cheaper of using original bound and changing the exit bound.
2140 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2142         PR tree-optimization/29516
2143         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2144         most_expensive_mult_to_index, addr_to_parts,
2145         create_mem_ref, maybe_fold_tmr): Make the type of
2146         fields of TARGET_MEM_REF sizetype.
2147         (move_fixed_address_to_symbol, move_pointer_to_base):
2148         New functions.
2149         * tree.def (TARGET_MEM_REF): Add comment on types of
2150         the operands.
2152 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
2154         * c-common.c (vector_types_convertible_p): Treat opaque types as
2155         always convertible if they have the same size, but not otherwise.
2157 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
2159         * ifcvt.c (struct noce_if_info): Add comments to the fields.
2160         Remove the b_unconditional field.
2161         (noce_try_sign_mask): Do not look at b_unconditional.
2162         (noce_process_if_block): Do not use merge_if_blocks.  Update
2163         the CFG here.  Do not set b_unconditional.
2164         (cond_move_process_if_block): Likewise.
2165         (find_cond_trap): Likewise.
2166         (check_cond_move_block): Require simple jump insns at the end
2167         of the basic block.
2169 2007-01-11  Jan Hubicka  <jh@suse.cz>
2171         PR tree-optimization/1046
2172         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2173         when alias info is not ready.
2174         (pass_tail_recursion): Do not require aliasing.
2175         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2176         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2177         * tree-ssa-copy.c (pass_copy_prop): Likewise.
2178         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2179         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2180         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2181         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2182         before inlining.
2183         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2184         When aliasing is not build, mark statement as volatile.
2186 2007-01-11  Tom Tromey  <tromey@redhat.com>
2188         PR preprocessor/15185, PR preprocessor/20989:
2189         * doc/cppopts.texi <-MT>: Update description of algorithm for
2190         computing default target.
2191         <-M, -MD>: Reword "basename" text.
2193 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2195         * builtins.c (expand_builtin_pow, expand_builtin_powi,
2196         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2197         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2198         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2199         fold_builtin_bswap, real_constp, fold_builtin_pow,
2200         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2201         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2202         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2203         * convert.c (convert_to_pointer): Likewise.
2204         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2205         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2206         const_binop, fold_convert_const_int_from_int,
2207         fold_convert_const_int_from_real,
2208         fold_convert_const_real_from_real, sign_bit_p,
2209         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2210         fold_sign_changed_comparison, fold_unary, fold_comparison,
2211         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2212         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2213         * print-tree.c (print_node_brief, print_node): Likewise.
2214         * stor-layout.c (place_field, layout_type): Likewise.
2215         * tree-chrec.c (keep_cast): Likewise.
2216         * tree.c (build_vector, build_real, build_real_from_int_cst,
2217         build_complex): Likewise.
2219 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2221         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
2222         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2224 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
2226         * configure.ac (strict1_warn): Rename to strict_warn.
2227         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2228         cc_set_by_configure, quoted_cc_set_by_configure,
2229         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2230         all_boot_languages, all_stagestuff): Remove.
2231         (target_list): Remove bootstrap targets.
2232         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2233         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2234         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2235         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2236         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2237         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2238         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2239         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2240         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2241         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2242         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2243         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2244         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2245         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2246         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2247         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2248         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2249         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2250         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2251         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2252         (mostlyclean): Adjust.
2253         (clean, distclean): Don't mention bootstrap stuff.
2254         * configure: Regenerate.
2255         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2256         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2257         treelang/config-lang.in (stagestuff): Remove.
2258         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2259         stageprofile, stagefeedback, stagestuff): Remove mention.
2261 2007-01-11  Nick Clifton  <nickc@redhat.com>
2263         * config/mcore/predicates.md (mcore_general_movesrc_operand):
2264         Accept CONSTs.
2265         (mcore_general_movdst_operand): Do not accept CONST_INTs.
2266         (mcore_arith_K_S_operand): Run the test for the S constraint not
2267         the test for the M constraint.
2268         (mcore_addsub_operand): Do not accept integer values that are
2269         larger than 32 bits.
2270         * config/mcore/mcore.md: Remove unused constraints from split.
2271         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2272         (addsi3): Likewise.
2273         (allocate_stack): Likewise.
2274         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2275         operands to 32 bits.
2276         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2277         INTVAL.
2278         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2279         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2280         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2281         mcore_output_bclri, mcore_output_andn, output_inline_const,
2282         mcore_output_move, mcore_output_movedouble): Likewise.
2283         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2284         (output_inline_const): Likewise.
2285         (output_inline_const): Fix format strings used in sprintf
2286         statements.
2287         * config/mcore/mcore-protos.h: Update prototypes for changed
2288         functions in mcore.c.
2289         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2290         HOST_WIDE_INT and not int.
2291         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2292         CONST_OK_FOR_N): Likewise.
2293         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2294         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2295         an INTVAL.
2297 2007-01-10  Jan Hubicka  <jh@suse.cz>
2299         * tree-vrp.c (remove_range_assertions): Release defs.
2300         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2301         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2303 2007-01-10  Paul Brook  <paul@codesourcery.com>
2305         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2306         * config/arm/arm.md (mulsi3subsi): New insn.
2308 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
2310         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2311         (tree_transform_and_unroll_loop): New.
2312         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2313         Declare.
2315 2007-01-10  Robert Kennedy <jimbob@google.com>
2317         * fold-const.c (fold_comparison): Fold comparisons like (x *
2318         1000 < 0) to (x < 0).
2320 2007-01-10  Ian Lance Taylor  <iant@google.com>
2322         * tree-pretty-print.c (dump_generic_node): Print parentheses when
2323         operands have the same priority.
2325 2007-01-10  Tom Tromey  <tromey@redhat.com>
2327         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2328         (fold_binary): Likewise.
2329         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2330         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2331         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2332         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2333         * langhooks.h (struct lang_hooks): Removed field
2334         'can_use_bit_fields_p'.
2336 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
2338         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2340 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
2342         * function.c (get_last_funcdef_no): New function.
2343         * function.h (get_last_funcdef_no): Declare.
2344         * tree-inline.c (initialize_cfun): Add initialization.
2345         (tree_function_versioning): Cleanup.
2347 2007-01-10  Jan Hubicka  <jh@suse.cz>
2349         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2350         PHIs.
2352 2007-01-10  Sa Liu  <saliu@de.ibm.com>
2353             Ben Elliston  <bje@au.ibm.com>
2355         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2356         for SAVE_NONLOCAL and Pmode for any other save level.
2357         * spu-protos.h (spu_restore_stack_block): Declare.
2358         * spu.md (save_stack_block): Remove.
2359         (restore_stack_block): Call spu_restore_stack_block.
2360         * spu.c (spu_restore_stack_block): New function.
2361         (spu_expand_epilogue): Remove old comment.
2363 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
2365         PR tree-optimization/30322
2366         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2367         (cand_value_at): Return the value as aff_tree.
2368         (may_eliminate_iv): Convert the bound from aff_tree to tree.
2369         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2370         aff_combination_mult): New functions.
2371         (aff_combination_add): Use aff_combination_add_cst.
2372         (aff_combination_convert): Allow conversions to a wider type.
2373         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2374         * tree-affine.h (aff_combination_mult): Declare.
2376 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
2378         * doc/tm.texi: Update documentation to reflect reality of exec
2379         and start file search behaviours. Update copyright year.
2380         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2381         header file directories.
2383 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
2385         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2386         (*cosxf2): Rename to cosxf2_i387.
2387         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
2388         using X87MODEF12 mode macro. Rename patterns to
2389         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
2390         for SSE math.
2391         (*cosdf2, *cossf2): Ditto.
2392         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
2393         to match extended input operands.
2394         (sincos<mode>3): New expander.
2395         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2396         insn patterns and corresponding splitters.
2398 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
2400         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2401         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2403 2007-01-09  Nicolas Pitre  <nico@cam.org>
2405         PR target/30173
2406         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2408 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2410         * target.h (struct gcc_target): New field library_rtti_comdat.
2411         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2412         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2413         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2414         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2416 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2418         * doc/invoke.texi (Optimize Options): Correct description of -O0.
2420 2007-01-08  Richard Guenther  <rguenther@suse.de>
2422         * tree.h (force_fit_type_double): Export.
2423         (force_fit_type): Remove.
2424         * fold-const.c (force_fit_type_double): New function.
2425         (force_fit_type): Remove.
2426         (int_const_binop): Use it.
2427         (fold_convert_const_int_from_int): Likewise.
2428         (fold_convert_const_int_from_real): Likewise.
2429         (fold_div_compare): Likewise.
2430         (fold_sign_changed_comparison): Likewise.
2431         (fold_unary): Likewise.
2432         (fold_negate_const): Likewise.
2433         (fold_abs_const): Likewise.
2434         (fold_not_const): Likewise.
2435         * c-common.c (shorten_compare): Use force_fit_type_double.
2436         * convert.c (convert_to_pointer): Likewise.
2438 2007-01-08  Richard Guenther  <rguenther@suse.de>
2440         * tree.h (build_int_cst_wide_type): Export.
2441         * tree.c (build_int_cst_wide_type): New function.
2442         (build_int_cst_wide): Fix comment.
2443         * builtins.c (fold_builtin_object_size): Use build_int_cst
2444         to build -1 or 0 of the correct type.  Use fit_double_type
2445         to check for overflow.
2446         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2447         to build the mask.
2448         (decode_field_reference): Likewise.
2449         (all_ones_mask_p): Likewise.
2450         (native_interpret_int): Use build_int_cst_wide_type.
2451         (fold_binary): Use build_int_cst_type to build an all-ones
2452         value.
2453         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2455 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
2457         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2459 2007-01-09  Ben Elliston  <bje@au.ibm.com>
2461         * genautomata.c (STATS_OPTION): New option.
2462         (stats_flag): New flag.
2463         (gen_automata_option): Handle it.
2464         (initiate_automaton_gen): Ditto.
2465         (write_automata): Output statistics only if stats_flag is
2466         set. Likewise, output time statistics only if time_flag is set.
2467         * doc/md.texi (Processor pipeline description): Document new flag.
2469 2007-01-08  Richard Guenther  <rguenther@suse.de>
2471         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2472         * tree.c (build_int_cst_type): Likewise.
2473         (size_in_bytes): Don't call force_fit_type on the result.
2474         (int_fits_type_p): Use fit_double_type.
2475         * fold-const.c (fit_double_type): New function.
2476         (force_fit_type): Use it.
2477         * tree.h (fit_double_type): Export.
2479 2007-01-08  Jan Hubicka  <jh@suse.cz>
2481         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2482         * ipa.c (function_and_variable_visibility): Fix return type.
2484 2007-01-08  Richard Guenther  <rguenther@suse.de>
2486         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2487         of offset to build the index.
2488         * tree-pretty-print.c (dump_generic_node): Don't build negated
2489         const just for printing.
2490         * c-pretty-print.c (pp_c_integer_constant): Likewise.
2491         * builtins.c (fold_builtin_int_roundingfn): Check if result
2492         fits the type by using force_fit_type and comparing the result.
2493         * predict.c (predict_loops): Use compare_tree_int for comparison.
2494         * tree.c (build_int_cst): Fall back to integer_type_node for
2495         NULL_TREE type.
2496         (build_int_cst_wide): Assert type is non-null.
2498 2007-01-08  Roberto Costa  <roberto.costa@st.com>
2500         * tree-vrp.c (extract_range_from_cond_expr): New.
2501         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2502         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2503         as expressions.
2504         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2505         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2506         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2507         used as expressions.
2508         * tree-object-size.c (cond_expr_object_size): New.
2509         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2511 2007-01-08  Jan Hubicka  <jh@suse.cz>
2513         * tree-ssa-forwprop.c (forward_propagate_into_cond,
2514         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2515         statement.
2517 2007-01-08  Richard Guenther  <rguenther@suse.de>
2519         PR tree-optimization/23603
2520         * tree-vrp.c (set_value_range_to_truthvalue): New function.
2521         (extract_range_from_binary): Fall back to truthvalue instead of
2522         varying for TRUTH_*_EXPR.
2523         (extract_range_from_comparison): Fall back to truthvalue instead of
2524         varying.
2525         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2526         if all visited PHI values were constant.
2528 2007-01-08  Jan Hubicka  <jh@suse.cz>
2530         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2531         (cgraph_analyze_function): break out from ...
2532         (cgraph_finalize_compilation_unit): ... here.
2533         (cgraph_expand_function): Remove forgoten commented out line.
2534         (cgraph_optimize): Analyze functions.
2536 2007-01-08  Jan Hubicka  <jh@suse.cz>
2538         * tree-pas.h (TODO_remove_function): New flag.
2539         (TODO_update*): Renumber.
2540         (pass_ipa_increase_alignment,
2541         pass_ipa_function_and_variable_visibility): New passes.
2542         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2543         (cgraph_function_and_variable_visibility): Move to ipa.c
2544         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2545         cgraph_increase_alignment.
2546         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2547         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2548         calling tree-inline.
2549         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2550         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2551         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2552         (gate_increase_alignment): New function.
2553         (pass_ipa_increase_alignment): New pass.
2554         * ipa.c: Inline tree-pass.h and timevar.h
2555         (function_and_variable_visibility): Move here from cgraphunit.c
2556         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2557         * passes.c (init_optimization_passes): Add the two new passes.
2558         (execute_todo): Handle cgraph_remove_functions.
2560 2007-01-08  Nick Clifton  <nickc@redhat.com>
2562         * config/frv/predicates.md (reg_or_0_operand): Accept
2563         CONST_DOUBLEs.
2565 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
2567         * config/bfin/rtems.h, config/bfin/t-rtems: New.
2568         * config.gcc: Add bfin*-rtems*.
2570 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2572         * c.opt: Add -flax-vector-conversions.
2573         * c-typeck.c (convert_for_assignment): Pass flag to
2574         vector_types_convertible_p to allow emission of note.
2575         (digest_init): Likewise.
2576         * c-opts.c: Handle -flax-vector-conversions.
2577         * c-common.c (flag_lax_vector_conversions): New.
2578         (vector_types_convertible_p): Unless -flax-vector conversions
2579         has been passed, disallow conversions between vectors with
2580         differing numbers of subparts and/or element types.  If such
2581         a conversion is disallowed, possibly emit a note on the first
2582         occasion only to inform the user of -flax-vector-conversions.
2583         The new last argument specifies this.
2584         * c-common.h (flag_lax_vector_conversions): New.
2585         (vector_types_convertible_p): Add extra argument.
2586         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2587         char_type_node for V*QI type vectors.
2588         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2589         Update to satisfy new typechecking rules.
2590         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2591         C and C++ variants.
2592         * doc/invoke.texi (C Dialect Options): Document
2593         -flax-vector-conversions.
2595 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2597         PR tree-optimization/29877
2598         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2599         a register variable on the RHS to not be replaceable.
2601 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
2602         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2603         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2604         const_simm15, const_pow2, const_npow2): Added.
2605         * config/score/misc.md (insv, extv, extzv, movmemsi,
2606         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2607         * config/score/score.c (score_address_cost, score_select_cc_mode):
2608         Added.
2609         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2610         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2611         Update score_print_operand.
2612         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2613         Adjust register allocate order and update some macro define.
2614         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2615         mdx_block_move_straight, mdx_block_move_loop_head,
2616         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2617         mdx_block_move): Added.
2618         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2619         mdp_limm): Updated and fix some bug and typo.
2620         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2621         (movsf, movdf, doloop_end): Added.
2623 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
2625         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2626         config/arm/thumb2.md: Fix comment typos.
2627         * doc/extend.texi: Fix a typo.
2629 2007-01-07  Eric Christopher  <echristo@apple.com>
2631         * configure.ac: Check for __stack_chk_fail for darwin.
2632         * configure: Regenerate.
2634 2007-01-07  Richard Guenther  <rguenther@suse.de>
2636         * tree-vrp.c (extract_range_from_assert): CSE calls to
2637         compare_values where possible.
2638         (extract_range_from_unary_expr): Likewise.
2640 2007-01-07  Anatoly Sokolov <aesok@post.ru>
2642         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2643         * config/avr/avr.c (call_insn_operand): Delete function.
2644         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2645         instead of register number. Use predicates.
2646         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2647         call_insn_operand): Add.
2649 2007-01-06  Jan Hubicka  <jh@suse.cz>
2651         * tree-pass.h (pass_build_cgraph_edges): Declare.
2652         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2653         (visited_nodes): Remove.
2654         (cgraph_create_edges): Move to cgraphbuild.c; rename to
2655         build_cgrpah_edges; make visited_nodes local.
2656         (cgraph_process_new_functions): DO not call initialize_inline_failed.
2657         (record_references_in_initializer): Move to cgraphbuild.c
2658         (initialize_inline_failed, rebuild_cgraph_edges,
2659         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2660         (verify_cgraph_node): Make visited_nodes local.
2661         (cgraph_analyze_function): Do not call cgraph_create_edges and
2662         initialize_inline_failed.
2663         (cgraph_expand_function): Do not call cgraph_lower_function;
2664         assert that function is already lowered.
2665         * Makefile.in (cgraphbuild.o): New.
2666         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2667         at the end of lowering passes.
2669 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
2671         * ifcvt.c (cond_move_convert_if_block): New function, code
2672         factored out from...
2673         (cond_move_process_if_block): ...here.  Call the new function
2674         on the THEN and ELSE blocks.
2675         (merge_if_block): Do not copy global_live_at_end, merge_blocks
2676         already takes care of this.
2678 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2680         PR c/19978
2681         * tree.h (TREE_OVERFLOW_P): New.
2682         * c-typeck.c (parser_build_unary_op): Warn only if result
2683         overflowed and operands did not.
2684         (parser_build_binary_op): Likewise.
2685         (convert_for_assignment): Remove redundant overflow_warning.
2686         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2688 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2690         * c-typeck.c (store_init_value): Split over two lines to follow
2691         the GNU coding style.
2693 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
2695         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2696         __GXX_EXPERIMENTAL_CXX0X__.
2697         * doc/cpp.texi: Same.
2699 2007-01-05  Richard Guenther  <rguenther@suse.de>
2701         PR middle-end/27826
2702         * tree.c (get_narrower): Do not construct COMPONENT_REFs
2703         with mismatched types.  Instead explicitly build a
2704         conversion NOP_EXPR.
2706 2007-01-05  Ian Lance Taylor  <iant@google.com>
2708         * c-common.c (decl_with_nonnull_addr_p): New function.
2709         (c_common_truthvalue_conversion): Call it.
2710         * c-typeck.c (build_binary_op): Likewise.
2711         * c-common.h (decl_with_nonnull_addr_p): Declare.
2713 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
2715         PR c/30360
2716         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2717         denom against 0.0.
2719 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
2721         * doc/install.texi (Final install): Document the fact that
2722         the GNAT runtime should not be stripped.
2724 2007-01-04  Jan Hubicka  <jh@suse.cz>
2726         * tree-inline.c (fold_marked_statements): Update operand caches
2727         and EH after folding
2729 2007-01-04  Ian Lance Taylor  <iant@google.com>
2731         * c-common.c (check_function_nonnull): Whitespace fix.
2733 2007-01-04  Jan Hubicka  <jh@suse.cz>
2735         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2736         comitted older version of patch.
2737         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2738         function, celanup cfg and collect garbage.
2740 2007-01-04  Mike Stump  <mrs@apple.com>
2742         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2743         (clean): Likewise.
2745 2007-01-04  Eric Christopher  <echristo@apple.com>
2747         * libgcc2.c (__bswapsi2): Use SItype.
2748         (__bswapdi2): Use DItype.
2749         * libgcc2.h: Update for above.
2751 2007-01-04  Paul Brook  <paul@codesourcery.com>
2753         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2754         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2755         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2756         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2757         umulsi3_highpart): Make conditional on !arm_arch6.
2758         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2759         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2760         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2761         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2762         umulsi3_highpart_v6): New insns.
2764 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2766         * fold-const.c (fold_convert): When casting an expression to void,
2767         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2768         have a type.  Instead of attempting to build a NOP_EXPR, return
2769         these "special" trees directly.
2771 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2773         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2774         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2775         rs6000_cost->fp not 0 as adjustment for outer NEG.
2777 2007-01-04  Jan Hubicka  <jh@suse.cz>
2779         * cgraph.c (cgraph_release_function_body): New function.
2780         (cgraph_remove_node): Use it.
2781         * cgraph.h (cgraph_release_function_body): Declare.
2782         * cgraphunit.c (cgraph_expand_function): Use it.
2783         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2784         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2785         is initialized and while compilation of other function is running.
2786         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2787         statement CFG annotations from here to ...
2788         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2790 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2792         * cfgloop.h (enum li_flags): Make the constants powers of two.
2794 2007-01-04  Jan Hubicka  <jh@suse.cz>
2796         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2797         set.
2798         (fold_marked_statements): New function.
2799         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2800         * tree-inline.h (copy_body_data): Add statements_to_fold.
2802 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2804         * config.gcc: Mention libgcc/config.host.
2805         * Makefile.in: Update comments mentioning libgcc.
2806         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2807         (all.cross, start.encap, rest.encap, rest.cross): Update
2808         dependencies for libgcc move.
2809         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2810         (clean-target, clean-target-libgcc): Delete.
2811         (srcdirify, GCC_EXTRA_PARTS): New macros.
2812         (libgcc-support, libgcc.mvars): New rules.
2813         (distclean): Remove mention of mklibgcc.
2814         (install): Don't reference INSTALL_LIBGCC.
2815         (install-common): Don't reference EXTRA_PARTS.
2816         (install-libgcc, install-multilib): Delete rules.
2817         * mklibgcc.in: Delete file.
2818         * doc/configfiles.texi: Don't mention mklibgcc.
2820         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2821         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2822         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2823         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2824         (mklibgcc, ldblspecs): Likewise.
2826         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2827         (posixpre.def): Use $(T).
2828         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2829         (SHLIB_SRC, SHLIB_INSTALL): Delete.
2830         (SHLIB_LINK): Make dummy.
2831         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2833         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2835         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2836         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2837         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2838         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2840         * configure.ac (all_outputs): Remove mklibgcc.
2841         * configure: Regenerated.
2843 2007-01-03  Josh Conner  <jconner@apple.com>
2845         PR middle-end/29683
2846         * calls.c (compute_argument_addresses): Set stack and stack_slot
2847         for partial args, too.
2848         (store_one_arg): Use locate.size.constant for the size when
2849         generating a save_area.
2851 2007-01-03  Robert Kennedy <jimbob@google.com>
2853         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2854         whose definitions are deleted due to basic block merging.
2856 2007-01-03  Paul Brook  <paul@codesourcery.com>
2858         PR target/16634
2859         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2860         functions.
2861         (use_return_insn): Return 0 for Thumb interrupt functions.
2862         (print_multi_reg): Add rfe argument for IRQ returns.
2863         (arm_output_epilogue): Pop interrupt return address directly into PC.
2864         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2866 2007-01-03  Paul Brook  <paul@codesourcery.com>
2868         Merge from sourcerygxx-4_1.
2869         * config/arm/thumb2.md: New file.
2870         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2871         Thumb-2.
2872         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2873         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2874         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2875         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2876         tables.
2877         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2878         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2879         comptibility.
2880         * config/arm/ieee754-sf.S: Ditto.
2881         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2882         (thumb1_base_register_rtx_p): ... to this.
2883         (thumb_index_register_rtx_p): Rename...
2884         (thumb1_index_register_rtx_p): ... to this.
2885         (thumb_output_function_prologue): Rename...
2886         (thumb1_output_function_prologue): ... to this.
2887         (thumb_legitimate_address_p): Rename...
2888         (thumb1_legitimate_address_p): ... to this.
2889         (thumb_rtx_costs): Rename...
2890         (thumb1_rtx_costs): ... to this.
2891         (thumb_compute_save_reg_mask): Rename...
2892         (thumb1_compute_save_reg_mask): ... to this.
2893         (thumb_final_prescan_insn): Rename...
2894         (thumb1_final_prescan_insn): ... to this.
2895         (thumb_expand_epilogue): Rename...
2896         (thumb1_expand_epilogue): ... to this.
2897         (arm_unwind_emit_stm): Rename...
2898         (arm_unwind_emit_sequence): ... to this.
2899         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2900         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2901         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2902         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2903         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2904         thumb2_output_casesi): New functions.
2905         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2906         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2907         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2908         THUMB2_WORK_REGS): Define.
2909         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2910         arm_condexec_mask, arm_condexec_masklen)): New variables.
2911         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2912         (arm_override_options): Check new CPU capabilities.
2913         Set new architecture flag variables.
2914         (arm_isr_value): Handle v7m interrupt functions.
2915         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
2916         Thumb-2.
2917         (const_ok_for_arm): Handle Thumb-2 constants.
2918         (arm_gen_constant): Ditto.  Use movw when available.
2919         (arm_function_ok_for_sibcall): Return false for v7m interrupt
2920         functions.
2921         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2922         (thumb_find_work_register, arm_load_pic_register,
2923         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2924         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2925         print_multi_reg, output_mov_long_double_fpa_from_arm,
2926         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2927         output_mov_double_fpa_from_arm, output_move_double,
2928         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2929         output_return_instruction, arm_output_function_prologue,
2930         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2931         arm_output_mi_thunk, thumb_set_return_address): Ditto.
2932         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
2933         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2934         (arithmetic_instr, shift_op): Use arm_shift_nmem.
2935         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
2936         '!' and 'L'.
2937         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2938         (thumb_expand_prologue): Use thumb_set_frame_pointer.
2939         (arm_file_start): Output directive for unified syntax.
2940         (arm_unwind_emit_set): Handle stack alignment instruction.
2941         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2942         Add v6t2, v7, v7a, v7r and v7m.
2943         (RETLDM): Add Thumb-2 code.
2944         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2945         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2946         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2947         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2948         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2949         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2950         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2951         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2952         ADJUST_INSN_LENGTH): Define.
2953         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2954         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2955         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2956         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2957         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2958         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2959         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2960         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2961         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2962         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2963         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2964         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2965         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2966         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2967         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2968         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2969         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2970         cortex-m3.
2971         * config/arm/arm-tune.md: Regenerate.
2972         * config/arm/arm-protos.h: Update prototypes.
2973         * config/arm/vfp.md: Enable patterns for Thumb-2.
2974         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
2975         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2976         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2977         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2978         * config/arm/libunwind.S: Add Thumb-2 code.
2979         * config/arm/constraints.md: Update include Thumb-2.
2980         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2981         * config/arm/ieee754-df.S: Ditto.
2982         * config/arm/bpabi.S: Ditto.
2983         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2984         * config/arm/predicates.md (low_register_operand,
2985         low_reg_or_int_operand, thumb_16bit_operator): New.
2986         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2987         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2988         * config/arm/t-arm-elf: Add armv7 multilib.
2989         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2990         Include thumb2.md.
2991         (UNSPEC_STACK_ALIGN, ce_count): New.
2992         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2993         arm_zero_extendsidi2, arm_zero_extendqidi2): New
2994         insns/expanders.
2995         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2996         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2997         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2998         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2999         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3000         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3001         insns.
3002         * doc/extend.texi: Document ARMv7-M interrupt functions.
3003         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3005 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
3007         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3008         (struct _Unwind_Context): Rename args_size to flags, remove
3009         signal_frame field, add a new args_size field and version field.
3010         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3011         _Unwind_IsExtendedContext): New inline functions.
3012         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3013         Assume by_value array is only present if _Unwind_IsExtendedContext.
3014         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3015         _Unwind_IsSignalFrame.
3016         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3017         (uw_update_context_1): Use _Unwind_SetSignalFrame.
3018         (uw_init_context_1): Initialize context->flags to
3019         EXTENDED_CONTEXT_BIT.
3020         * config/rs6000/linux-unwind.h (frob_update_context): Use
3021         _Unwind_SetSignalFrame.
3023 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3025         PR middle-end/30353
3026         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3027         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3028         the value.
3030 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3032         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3033         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3034         RET, AT_SP): Remove.
3035         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3036         instead of RET.
3038         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3039         * cfgbuild.c (find_basic_blocks): Likewise.
3040         * cfgrtl.c (rtl_create_basic_block): Likewise.
3041         * function.c (temp_slots_at_level): Likewise.
3042         * reg-stack.c (stack_regs_mentioned): Likewise.
3043         * regclass.c (allocate_reg_info): Likewise.
3044         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3045         set_bb_for_stmt, move_block_to_fn): Likewise.
3046         * tree-complex.c (tree_lower_complex): Likewise.
3047         * vec.h (VEC_safe_grow_cleared): New.
3049         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3051 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
3053         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3054         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3055         to loop_version.
3056         * cfgloopmanip.c (scale_loop_frequencies): Export.
3057         (loopify): Scale the frequencies by prescribed coefficients.
3058         (set_zero_probability): New function.
3059         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3060         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3061         and frequencies according to arguments.
3062         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3063         correctly.
3064         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3065         * modulo-sched.c (sms_schedule): Set probabilities for entering
3066         versioned loop correctly.
3067         * tree-vect-transform.c (vect_transform_loop): Ditto.
3068         * cfgloop.h (loopify, loop_version): Declaration changed.
3069         (scale_loop_frequencies): Declared.
3071 2007-01-02  Jan Hubicka  <jh@suse.cz>
3073         * cgraph.c: Include tree-flow.h
3074         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3075         early_local_passes.
3076         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3077         * tree-pass.h (pass_all_early_optimizations): Declare.
3078         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3079         early_local_passes.
3080         (cgraph_analyze_function): Do early_local_passes.
3081         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3082         Do not add referenced vars.
3083         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3084         SSA form.
3085         (gate_all_early_local_passes): New gate.
3086         (pass_early_local_passes): Use new gate.
3087         (execute_early_local_optimizations): New functions.
3088         (gate_all_early_optimizations): New gate.
3089         (pass_all_early_optimizations): New pass.
3090         (execute_free_datastructures): Free SSA only when initialized.
3091         (gate_init_datastructures): Init only when optimizing.
3092         (tree_lowering_passes): Do early local passes when called late.
3093         * tree-profile.c (do_tree_profiling): Don't profile functions added
3094         late.
3095         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3096         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3097         not active.
3098         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3099         during early_local_passes.
3100         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3103 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3105         * Makefile.in: Update copyright year.
3107 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3109         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3110         in $(lang_checks) and check-consistency targets.
3112 2007-01-02  Jan Hubicka  <jh@suse.cz>
3114         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3115         Do not add referenced vars.
3116         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3117         are not active.
3118         * passes.c (init_optimization_passes): Put mudflap_2 after
3119         free_datastructures.
3121 2007-01-02  Jan Hubicka  <jh@suse.cz>
3123         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3124         Set NOTHROW flag on call statements proved to be nothrow.
3125         Update statement of local calls so new pure/const functions are
3126         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3127         goto receivers.
3128         (tree_rest_of_compilation): Register hooks and initialize bitmap
3129         early. Do not set after_inlining flag.
3131 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3133         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3134         HOST_BITS_PER_LONGLONG
3136 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3138         PR c/19977
3139         * c-typeck.c (store_init_value): Don't emit pedantic overflow
3140         warning for non-static initializers.
3142 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
3144         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3145         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3146         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3147         Always use set_unique_reg_note to add REG_EQUAL notes.
3149 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3151         Revert:
3152         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3154         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3155         * cfgbuild.c (find_basic_blocks): Likewise.
3156         * cfgrtl.c (rtl_create_basic_block): Likewise.
3157         * function.c (temp_slots_at_level): Likewise.
3158         * reg-stack.c (stack_regs_mentioned): Likewise.
3159         * regclass.c (allocate_reg_info): Likewise.
3160         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3161         set_bb_for_stmt, move_block_to_fn): Likewise.
3162         * tree-complex.c (tree_lower_complex): Likewise.
3163         * vec.h (VEC_safe_grow_cleared): New.
3165 2007-01-02  Ian Lance Taylor  <iant@google.com>
3167         * c-common.c (c_common_truthvalue_conversion): When warning about
3168         using an assignment as a truth value, set TREE_NO_WARNING.
3170 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3172         PR middle-end/7651
3173         * c.opt (Wold-style-declaration): New.
3174         * doc/invoke.texi (C-only Warning Options): New.
3175         (Wold-style-declaration): Document it.
3176         (Wextra): Enabled by -Wextra.
3177         * c-opts.c (c_common_post_options): Enabled by -Wextra.
3178         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3179         -Wold-style-declaration.
3181 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3183         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3184         * cfgbuild.c (find_basic_blocks): Likewise.
3185         * cfgrtl.c (rtl_create_basic_block): Likewise.
3186         * function.c (temp_slots_at_level): Likewise.
3187         * reg-stack.c (stack_regs_mentioned): Likewise.
3188         * regclass.c (allocate_reg_info): Likewise.
3189         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3190         set_bb_for_stmt, move_block_to_fn): Likewise.
3191         * tree-complex.c (tree_lower_complex): Likewise.
3192         * vec.h (VEC_safe_grow_cleared): New.
3194 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
3196         * c-common.c (c_common_nodes_and_builtins): Since variants of
3197         void_type_node get built before it is given a name, we need to
3198         give those variants the name, too.
3199         (complete_array_type): We need to work with the canonical main
3200         type of the array, from which we will build the qualified version.
3201         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3202         * print-tree.c (print_node): Display canonical type information
3203         for each type.
3204         * stor-layout.c (layout_type): When we don't know the
3205         alignment of a type for which we're building an array, we end up
3206         guessing wrong, so make the type require structural equality.
3207         * tree.c (make_node_stat): When we build a new type, it is its
3208         own canonical type.
3209         (build_type_attribute_qual_variant): When building an attribute
3210         variant, its canonical type is the non-attribute variant. However,
3211         if the attributes are target-dependent and they differ, we need to
3212         use structural equality checks for this type.
3213         (build_qualified_type): A qualified type is not equivalent to its
3214         unqualified variant; set the canonical type appropriately.
3215         (build_distinct_type_copy): When building a distinct type from
3216         another type, the new type is its own canonical type.
3217         (build_variant_type_copy): When building a new type variant, we
3218         assume that it is equivalent to the original type.
3219         (build_pointer_type_for_mode): When building a pointer type, also
3220         build a canonical type pointer.
3221         (build_reference_type_for_mode): When building a reference type,
3222         also build a canonical type reference.
3223         (build_index_type): When we can't hash an index type (e.g.,
3224         because its maximum value is negative), the index type requires
3225         structural equality tests.
3226         (build_array_type): Build the canonical form of an array type.
3227         (build_function_type): Function types require structural equality,
3228         because they contain default arguments, attributes, etc.
3229         (build_method_type_directly): Ditto for method types.
3230         (build_offset_type): Build the canonical offset type.
3231         (build_complex_type): Build the canonical vector type.
3232         (make_vector_type): Build the canonical vector type.
3233         * tree.h (TYPE_CANONICAL): New.
3234         (TYPE_STRUCTURAL_EQUALITY_P): New.
3235         (SET_TYPE_STRUCTURAL_EQUALITY): New.
3236         (struct tree_type): Added "canonical" field.
3237         * params.h (VERIFY_CANONICAL_TYPES): New.
3238         * doc/c-tree.texi (TYPE_CANONICAL): Document.
3239         (TYPE_STRUCTURAL_EQUALITY_P): Document.
3240         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3241         * doc/invoke.texi (verify-canonical-types): Document --param
3242         parameter for verifying canonical types.
3244 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3246         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3247         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3248         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3249         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3250         tm_file.
3251         * config/rs6000/e500.h: New.
3252         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3253         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3254         Remove.
3255         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3256         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3257         TARGET_E500_DOUBLE): Remove.
3258         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3259         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3260         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3261         * config/rs6000/rs6000.c (rs6000_override_options): Use
3262         CHECK_E500_OPTIONS.
3264 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3266         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3267         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3268         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3269         rs6000_emit_cmove): Don't check TARGET_E500.
3270         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3271         sordered): Don't check TARGET_E500.
3273 2007-01-01  Eric Christopher  <echristo@apple.com>
3275         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3276         HARD_REGISTER_NUM_P.
3278 2007-01-01  Roger Sayle  <roger@eyesopen.com>
3280         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3281         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3282         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3283         eq/ne Y".
3285 2007-01-01  Mike Stump  <mrs@apple.com>
3287         * configure.ac: Remove support for building with Apple's gcc-3.1.
3289 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3291         PR middle-end/30311
3292         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3293         REG is a hard register.  Do not modify REG before calling
3294         subreg_nregs.
3295         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3296         multiple of the size of YMODE for certain lowpart cases.
3298 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3300         PR middle-end/30253
3301         * gimplify (voidify_wrapper_expr): Update for
3302         GIMPLE_MODIFY_STMT.
3304 2007-01-01  Andreas Schwab  <schwab@suse.de>
3306         PR target/29166
3307         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3308         save of BR0 in extra_spill_size instead of spill_size.
3309         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3310         area.
3311         (ia64_expand_epilogue): Restore BR0 from its new location.
3313 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3315         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3316         Use a temporary variable if the left hand side is not a gimple
3317         register.
3319 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3321         * gimplify.c (gimplify_return_expr): Make the temporary variable
3322         for the return expression, a gimple register variable.
3324 2007-01-01  Jan Hubicka  <jh@suse.cz>
3326         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3327         INSN_LIST.
3329 2007-01-01  Mike Stump  <mrs@apple.com>
3331         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3332         .literal16.
3333         * config/darwin.c (machopic_select_rtx_section): Use
3334         HAVE_GAS_LITERAL16.
3335         (darwin_mergeable_constant_section): Likewise.
3336         * configure: Regenerate.
3337         * config.in: Regenerate.
3339 2007-01-01  Jan Hubicka  <jh@suse.cz>
3340             Andrew Pinski  <pinskia@gmail.com>
3342         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3343         before starting IPA passes.