OSDN Git Service

* pa.md (tp_load): Correct mfctl instruction syntax.
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-02-01  Guy Martin  <gmsoft@gentoo.org>
2
3         * pa.md (tp_load): Correct mfctl instruction syntax.
4
5 2007-02-01  Geoffrey Keating  <geoffk@apple.com>
6
7         * config/rs6000/rs6000.c (rs6000_stack_info): Correct
8         altivec_padding_size calculation on AIX.  Improve comment, add
9         assert to verify that it's right.
10
11         * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
12         64-bit.
13
14 2007-2-01   Seongbae Park <seongbae.park@gmail.com>
15
16         PR inline-asm/28686
17         * global.c (global_alloc): Add mising initialization of
18         ELIMINABLE_REGSET.
19
20 2007-02-01  Roger Sayle  <roger@eyesopen.com>
21
22         * alias.c (init_alias_analysis): Correct whitespace.
23         * bb-reorder.c (fix_edges_for_rarely_executed_code,
24         partition_hot_cold_basic_blocks): Likewise.
25         * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
26         expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
27         maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
28         fold_builtin_snprintf_chk, fold_builtin_printf,
29         fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
30         do_mpfr_arg3, do_mpfr_sincos): Likewise.
31         * cfgcleanup.c (cleanup_cfg): Likewise.
32         * cfgexpand.c (tree_expand_cfg): Likewise.
33         * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
34         * function.c (get_next_funcdef_no): Likewise.
35         * gengtype.c (main): Likewise.
36         * genmodes.c (main): Likewise.
37         * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
38         * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
39         * ifcvt.c (noce_emit_move_insn): Likewise.
40         * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
41         Likewise.
42         * stor-layout.c (get_best_mode): Likewise.
43         * tree-ssa-loop-niter.c (get_val_for): Likewise.
44         * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
45         scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
46         perform_var_substitution, solve_graph): Likewise.
47         * tree-vrp.c (vrp_finalize): Likewise.
48
49 2007-02-01  Ian Lance Taylor  <iant@google.com>
50
51         * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
52         high part of a paradoxical subreg, return a constant zero.
53
54 2007-02-01  Zdenek Dvorak <dvorakz@suse.cz>
55
56         * toplev.c (lang_dependent_init): Call init_set_costs.
57         * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
58
59 2007-02-01  Richard Guenther  <rguenther@suse.de>
60
61         PR middle-end/30656
62         * fold-const.c (fold_negate_expr): Allow negating a
63         constant if overflow does not change.
64
65 2007-02-01  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
66
67         * doc/c-tree.texi (Expression trees): Improve markup.
68         * doc/tm.texi (Register Classes, Addressing Modes)
69         (Floating Point): Fix spacing after abbreviations.  Fix some
70         typos.
71
72 2007-02-01  Ben Elliston  <bje@au.ibm.com>
73
74         * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
75         * doc/contrib.texi: Likewise.
76         * doc/install.texi: Likewise.
77
78 2007-01-31  Richard Henderson  <rth@redhat.com>
79             Ian Lance Taylor  <iant@google.com>
80
81         * lower-subreg.c: New file.
82         * rtl.def (CONCATN): Define.
83         * passes.c (init_optimization_passes): Add pass_lower_subreg and
84         pass_lower_subreg2.
85         * emit-rtl.c (update_reg_offset): New static function, broken out
86         of gen_rtx_REG_offset.
87         (gen_rtx_REG_offset): Call update_reg_offset.
88         (gen_reg_rtx_offset): New function.
89         * regclass.c: Revert patch of 2006-03-05, restoring
90         reg_scan_update.
91         (clear_reg_info_regno): New function.
92         * dwarf2out.c (concatn_loc_descriptor): New static function.
93         (loc_descriptor): Handle CONCATN.
94         * common.opt (fsplit_wide_types): New option.
95         * opts.c (decode_options): Set flag_split_wide_types when
96         optimizing.
97         * timevar.def (TV_LOWER_SUBREG): Define.
98         * rtl.h (gen_reg_rtx_offset): Declare.
99         (reg_scan_update): Declare.
100         * regs.h (clear_reg_info_regno): Declare.
101         * tree-pass.h (pass_lower_subreg): Declare.
102         (pass_lower_subreg2): Declare.
103         * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
104         (Optimize Options): Add -fsplit-wide-types to -O1 list.  Document
105         -fsplit-wide-types.
106         * doc/rtl.texi (Regs and Memory): Document concat and concatn.
107         * Makefile.in (OBJS-common): Add lower-subreg.o.
108         (lower-subreg.o): New target.
109
110 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
111
112         * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
113
114 2007-01-31  Anatoly Sokolov <aesok@post.ru>
115
116         PR target/19087
117         * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
118
119 2007-01-31  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
120
121         PR middle-end/29335
122         * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
123
124 2007-01-31  Zdenek Dvorak <dvorakz@suse.cz>
125
126         * cfgloop.h: Include vec-prim.h.
127         (enum li_flags): Remove LI_ONLY_OLD.
128         (loop_iterator): Changed.
129         (fel_next, fel_init): Iterate over loop tree.
130         (FOR_EACH_LOOP_BREAK): New macro.
131         * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
132         FOR_EACH_LOOP.
133         * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
134         * modulo-sched.c (sms_schedule): Ditto.
135         * tree-vectorizer.c (vectorize_loops): Ditto.
136         * doc/loop.texi: Update information on loop numbering and behavior of
137         FOR_EACH_LOOP wrto new loops.
138         * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
139         add_to_evolution_1): Test nestedness of loops instead of comparing
140         their numbers.
141         * tree-chrec.c (chrec_fold_plus_poly_poly,
142         chrec_fold_multiply_poly_poly, chrec_evaluate,
143         hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
144         reset_evolution_in_loop): Ditto.
145         * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
146
147 2007-01-31  Dirk Mueller  <dmueller@suse.de>
148
149         * c-common.c (warn_about_parentheses): Separate warning about
150         un-parenthized sequence of comparison operators from the one
151         which is supposed to warn about x <= y <= z.
152
153 2007-01-31  Uros Bizjak  <ubizjak@gmail.com>
154
155         * optabs.h (enum optab_index): Add new OTI_isinf.
156         (isinf_optab): Define corresponding macro.
157         * optabs.c (init_optabs): Initialize isinf_optab.
158         * genopinit.c (optabs): Implement isinf_optab using isinf?f2
159         patterns.
160         * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
161         (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
162         using isinf_optab.
163         (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
164         expand_builtin_interclass_mathfn.
165         * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
166         * config/i386/i386.md (UNSPEC_FXAM): New constant.
167         (fxam<mode>2_i387): New insn pattern.
168         (isinf<mode>2) New expander to implement isinf, isinff and isinfl
169         built-in functions as x87 inline asm.
170
171 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
172
173         * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
174
175 2007-01-30  Eric Christopher  <echristo@apple.com>
176
177         * config.gcc: Add geode.
178
179 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
180
181         * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
182         ipa-inline.c, tree-profile.c, tree-ssa-live.c,
183         tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
184         value-prof.c: Fix comment typos.
185
186 2007-01-30  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
187
188         PR c++/24745
189         * doc/invoke.texi (Wpointer-arith): Document warning.
190
191 2007-01-30  Janis Johnson
192
193         * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
194         the list of discrepancies from the draft TR.
195
196 2007-01-30  Dirk Mueller  <dmueller@suse.de>
197
198         PR c++/30601
199         * doc/invoke.texi (-Wreturn-type): Update description to
200         match new behavior.
201
202 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
203
204         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
205
206 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
207
208         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
209         to handle optabs that operate on floating point input argument and
210         output to integer output.
211         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
212         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
213         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
214         expand_builtin_interclass_mathfn ().
215         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
216         (ilogbsi2): Remove.
217         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
218         ilogbl built-in functions as x87 intrinsics.
219
220 2007-01-30  Richard Guenther  <rguenther@suse.de>
221
222         PR middle-end/27657
223         * dwarf2out.c (reference_to_unused): Query varpool if the
224         variable was output.
225
226 2007-01-30  Richard Guenther  <rguenther@suse.de>
227
228         PR middle-end/30313
229         * passes.c (execute_one_pass): Reset in_gimple_form to not
230         confuse non-unit-at-a-time mode.
231
232 2007-01-29  Roger Sayle  <roger@eyesopen.com>
233             Richard Guenther  <rguenther@suse.de>
234
235         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
236         avoid undefined behaviour on overflow.  Use force_fit_type_double
237         to construct the constant with the specified TREE_OVERFLOW.
238
239 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
240
241         * config/dfp-bit.c: Add parameterized support for fp exceptions.
242         * config/dfp-bit.h: Ditto.
243
244 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
245
246         * c-decl.c (pop_scope): Replace warnings with call to
247         warn_for_unused_label.
248         * c-common.h (warn_for_unused_label): Declare.
249         * c-common.c (warn_for_unused_label): Define.
250
251 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
252
253         * tree-optimize.c (update_inlined_to_pointers): Delete unused
254         function.
255
256 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
257
258         * Makefile.in (USER_H): Remove decfloat.h.
259         * ginclude/decfloat.h: Delete, moving contents to ...
260         * ginclude/float.h: Add support for decimal floating point,
261         guarded by __STDC_WANT_DEC_FP__.
262
263 2007-01-29  Mike Stump  <mrs@apple.com>
264
265         * doc/gccint.texi (Top): Rename Loop Representation to Loop
266         Analysis and Representation to resolve case insensitive conflict.
267         * doc/loop.texi (Loop Analysis and Representation): Likewise.
268
269 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
270
271         * tree.h (struct tree_memory_tag): Add aliases member.
272         (MTAG_ALIASES): New macro.
273         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
274         (add_may_alias): Remove pointer-set. Update for may_aliases being
275         a bitmap.
276         (mark_aliases_call_clobbered): Update for may_aliases being a
277         bitmap.
278         (compute_tag_properties): Ditto.
279         (create_partition_for): Ditto.
280         (compute_memory_partitions): Ditto.
281         (dump_may_aliases_for): Ditto.
282         (is_aliased_with): Ditto.
283         (add_may_alias_for_new_tag): Ditto.
284         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
285         (compute_is_aliased): New function.
286         (compute_may_aliases): Call compute_is_aliased).
287         (init_alias_info): Initialize alias_bitmap_obstack.
288         (union_alias_set_into): New function.
289         (compute_flow_sensitive_aliasing): Use union_aliases_into.
290         (have_common_aliases_p): Rewrite to take two bitmaps and use
291         intersection.
292         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
293         Update for bitmaps.
294         (finalize_ref_all_pointers): Update for add_may_alias changes.
295         (new_type_alias): Ditto.
296         * tree-flow-inline.h (may_aliases): Return a bitmap.
297         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
298         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
299         may_aliases being a bitmap.
300         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
301         may_aliases now returns a bitmap.
302         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
303         being a bitmap.
304         * tree-ssa-operands.c (add_virtual_operand): Update for
305         may_aliases being a bitmap.
306
307 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
308
309         PR tree-optimization/30630
310         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
311         variable as changed.
312
313 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
314
315         PR c++/28266
316         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
317         expanded only once even if an error occurs.
318
319 2007-01-29  Ben Elliston  <bje@au.ibm.com>
320
321         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
322         (__gcov_average_profiler): Likewise.
323         (__gcov_ior_profiler): Likewise.
324         (__gcov_merge_ior): Likewise.
325
326 2007-01-28  Jan Hubicka  <jh@suse.cz>
327
328         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
329         * value-prof.c (stringop_block_profile): Likewise.
330
331 2007-01-28  Jan Hubicka  <jh@suse.cz>
332
333         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
334         variant handling histograms; add wrapper.
335         (clear_storage_via_libcall): Export.
336         (emit_block_move_hints): Break out from ...; add histograms.
337         (emit_block_move): ... this one.
338         (clear_storage_hints): Break out from ...; add histograms.
339         (clear_storage): ... this one.
340         (set_storage_via_memset): Handle histogram.
341         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
342         (clear_storage_hints, clear_storage_via_libcall): Declare.
343         (set_storage_via_setmem): Update prototype.
344         * doc/md.texi (movmem, setmem): Document new arguments.
345
346         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
347         new histograms.
348         (stringop_block_profile): New global function.
349         (tree_stringops_values_to_profile): Profile block size and alignment.
350         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
351         HIST_TYPE_IOR.
352         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
353         (stringop_block_profile): Declare.
354         * builtins.c: Include value-prof.h.
355         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
356         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
357         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
358         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
359         * profile.c (instrument_values): Add new counters.
360         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
361         calls.
362         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
363         (tree_init_edge_profiler): Build new profilers.
364         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
365         (pass_tree_profile): Add dump.
366         (tree_profile_hooks): Update.
367         * Makefile.in (LIBGCOV): Add new constants.
368         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
369         __gcov_ior_profiler): New.
370         * i386.md (movmem/setmem expanders): Add new optional arguments.
371
372 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
373
374         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
375
376 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
377
378         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
379         (tree_unswitch_single_loop): Pass eni_size_weights to
380         tree_num_loop_insns.
381         * tree-ssa-loop-manip.c: Include tree-inline.h.
382         (can_unroll_loop_p): Pass eni_size_weights to
383         tree_num_loop_insns.
384         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
385         Pass eni_size_weights to estimate_num_insns.
386         * tree.h (init_inline_once): Export.
387         * toplev.c (backend_init): Call init_inline_once.
388         * cgraphunit.c (cgraph_process_new_functions,
389         cgraph_analyze_function): Pass eni_inlining_weights to
390         estimate_num_insns.
391         * ipa-inline.c (compute_inline_parameters): Ditto.
392         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
393         to estimate_num_insns.
394         (try_unroll_loop_completely): Pass eni_size_weights to
395         tree_num_loop_insns.
396         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
397         ot estimate_num_insns.
398         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
399         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
400         * tree-inline.c (eni_inlining_weights, eni_size_weights,
401         eni_time_weights): New variables.
402         (init_inline_once): Initialize them.
403         (struct eni_data): Mew.
404         (estimate_num_insns_1, estimate_num_insns): Use weights.
405         * tree-inline.h (struct eni_weights_d): New.
406         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
407         (estimate_num_insns): Declaration changed.
408         * cfgloop.h (tree_num_loop_insns): Declaration changed.
409         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
410         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
411
412 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
413
414         * tree-data-ref.c (conflict_fn): Assert that the number of affine
415         relations in the conflict function is valid.
416
417 2007-01-27  Ian Lance Taylor  <iant@google.com>
418
419         * common.opt: Add fstrict-overflow.
420         * opts.c (decode_options): Set flag_strict_overflow if -O2.
421         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
422         (TYPE_OVERFLOW_UNDEFINED): Define.
423         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
424         Replace all uses.
425         * tree.h (TYPE_TRAP_SIGNED): Don't define.
426         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
427         (fold_negate_expr): Likewise.
428         (make_range): Likewise.
429         (extract_muldiv_1): Likewise.
430         (maybe_canonicalize_comparison): Likewise.
431         (fold_comparison): Likewise.
432         (fold_binary): Likewise.
433         (tree_expr_nonnegative_p): Likewise.
434         (tree_expr_nonzero_p): Likewise.
435         * tree-vrp.c (compare_values): Likewise.
436         (extract_range_from_binary_expr): Likewise.
437         (extract_range_from_unary_expr): Likewise.
438         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
439         Likewise.
440         (nowrap_type_p): Likewise.
441         * tree-scalar-evolution.c (simple_iv): Likewise.
442         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
443         (build_range_check): Likewise.
444         (extract_muldiv_1): Likewise.
445         (fold_comparison): Likewise.
446         * tree-vrp.c (vrp_int_const_binop): Likewise.
447         (extract_range_from_unary_expr): Likewise.
448         * convert.c (convert_to_integer): Likewise.
449         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
450         (fold_comparison): Likewise.
451         (fold_binary): Likewise.
452         * optabs.c (optab_for_tree_code): Likewise.
453         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
454         * simplify-rtx.c (simplify_const_relational_operation): Check
455         flag_strict_overflow and flag_trapv.
456         (simplify_const_relational_operation): Likewise.
457         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
458         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
459         -fstrict-overflow.
460
461 2007-01-27  Roger Sayle  <roger@eyesopen.com>
462
463         * tree.c (tree_fold_gcd): Delete.
464         * tree.h (tree_fold_gcd): Remove prototype.
465         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
466         test whether one constant integer is a multiple of another.  Instead
467         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
468         * fold-const.c (multiple_of_p):  We've determined both TOP and
469         BOTTOM are integer constants so we can call int_const_binop directly
470         instead of the more generic const_binop.
471
472 2007-01-27  Roger Sayle  <roger@eyesopen.com>
473
474         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
475         1*X check that the constant hasn't overflowed, to preserve the
476         TREE_OVERFLOW bit.
477         (round_up): Provide an efficient implementation when rouding-up an
478         INTEGER_CST to a power-of-two.
479
480 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
481
482         * doc/sourcebuild.texi: Add comma for clarity.
483         * doc/extend.texi: Fix some typos.
484         * doc/passes.texi: Likewise.
485         * doc/cppinternals.texi: Likewise.
486         * doc/c-tree.texi: Likewise.
487         * doc/tree-ssa.texi: Likewise.
488         * doc/install.texi: Likewise.
489
490 2007-01-27  Jan Hubicka  <jh@suse.cz>
491
492         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
493         (pass_sra): Do not require alias information.
494         * passes.c (init_optimization_passes): Add SRA
495
496 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
497
498         * tracer.c (rest_of_handle_tracer): We already cleaned
499         up the CFG in tracer() so don't do it here again.
500         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
501         cleanup_cfg here, either.  And don't call renumber_insns.
502
503         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
504         and BB_HEAD are in the insn stream here.  Instead make sure
505         that BB_INSN is valid on all insns.  Also, do check here that
506         there are no pending branch predictions...
507         (rtl_verify_flow_info): ...instead of doing it here.  Checks
508         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
509         here.
510
511 2007-01-26  Roger Sayle  <roger@eyesopen.com>
512
513         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
514         function to simplify/factorize operand order canonicalization.
515         (ix86_fixup_binary_operands): Reorganize using the above function.
516         (ix86_binary_operator_ok): Likewise.
517
518 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
519
520         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
521         file scope from optimize_attrs.
522         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
523         up search.
524         (optimize_attrs): Clear insn_code_values after freeing it.
525
526 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
527
528         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
529         a parts.base assignment.
530
531 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
532
533         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
534         (compute_subscript_distance, initialize_data_dependence_relation,
535         finalize_ddr_dependent, analyze_ziv_subscript,
536         analyze_siv_subscript_cst_affine,
537         compute_overlap_steps_for_affine_univar,
538         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
539         analyze_siv_subscript, analyze_miv_subscript,
540         analyze_overlapping_iterations, subscript_dependence_tester_1,
541         compute_self_dependence, free_dependence_relation): Work
542         with affine_fn instead of chrecs.
543         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
544         common_affine_function, affine_function_base,
545         affine_function_constant_p, affine_fn_op, affine_fn_plus,
546         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
547         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
548         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
549         (all_chrecs_equal_p): Removed.
550         * tree-data-ref.h (affine_fn, conflict_function): New types.
551         (struct subscript): Change type of conflicting_iterations_in_a
552         and conflicting_iterations_in_b.
553
554 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
555
556         PR other/30182
557         * config/pa/pa.h (TARGET_HPUX_11): New.
558         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
559         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
560
561 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
562         Richard Guenther  <rguenther@suse.de>
563
564         * tree-ssa-structalias.c (solve_graph): Handle case
565         we merged the variable to another.
566
567 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
568
569         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
570         functions.
571
572         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
573         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
574
575 2007-01-25  DJ Delorie  <dj@redhat.com>
576
577         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
578         allow changes to modes which don't fit in those registers.
579
580         * reload1.c (choose_reload_regs): Check for invalid subregs before
581         computing their locations, not after.
582
583 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
584
585         PR 25127
586         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
587         save Altivec registers in an eh_return function.
588         (compute_vrsave_mask): Likewise.
589         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
590         for saving Altivec registers.
591         (rs6000_emit_prologue): Don't allocate stack twice in
592         eh_return function.  Correct expected value of altivec_save_offset
593         when using save_world.  Describe save of R0 to stack when using
594         save_world.  Describe stack pointer adjustment when using
595         save_world.  Remove duplicated eh_return parameter register saving.
596         Update sp_offset variable after save_world.
597         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
598         darwin-world.asm.
599         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
600         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
601         implies Altivec.
602
603 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
604
605         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
606
607 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
608
609         PR other/30182
610         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
611         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
612
613 2007-01-25  Richard Guenther  <rguenther@suse.de>
614
615         * doc/invoke.texi (-Wcoverage-mismatch): Document.
616         * common.opt (-Wcoverage-mismatch): New warning option.
617         * coverage.c (get_coverage_counts): Ignore coverage mismatch
618         if -Wcoverage-mismatch is given.
619
620 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
621
622         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
623         (ipcp_driver): Change to static definition.
624         Add dumping of the ifunctions.
625         (constant_val_insert): Remove unused parameter. Support for SSA.
626         (ipcp_propagate_const): Support for SSA.
627         (ipcp_profile_bb_print): Print only analyzed nodes.
628         (ipcp_replace_map_create): Remove support for Fortran constant
629         for now.
630         * ipa-prop.c (ipa_method_modify_stmt,
631         ipa_callsite_compute_param): Support for SSA.
632         * ipa-prop.h (ipcp_driver): Remove declaration.
633         (IS_VALID_TREE_MAP_INDEX): Add define.
634
635 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
636
637         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
638
639 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
640
641         PR middle-end/30447
642         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
643         constant argument to cabs and do it without checking for
644         -funsafe-math-optimizations.
645
646 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
647
648         * c-common.h (RID_FIRST_CXX0X): New.
649         (RID_LAST_CXX0X): New.
650         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
651         by -Wall.
652         * c.opt (Wc++0x-compat): New.
653         * doc/invoke.texi (-Wc++0x-compat): Document.
654
655 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
656
657         * builtins.c (fold_builtin_carg): New.
658         (fold_builtin_1): Use it.
659
660 2007-01-24  Jan Hubicka  <jh@suse.cz>
661
662         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
663
664         * ipa-inline.c (initial_insns, max_insns): Delete.
665         (compute_max_insns): New function.
666         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
667         of insns as base for code growth.
668         (cgraph_decide_inlining): Make initial_insns local; do not compute
669         max_insns.
670         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
671         * doc/invoke.texi (inline-unit-growth): Update docs.
672
673 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
674
675         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
676         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
677         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
678         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
679         for CPUs that have PTA_CX16 set.
680
681         PR middle-end/27416
682         * gimplify.c (omp_check_private): New function.
683         (gimplify_scan_omp_clauses): Use it for
684         firstprivate/lastprivate/reduction.
685
686         PR middle-end/30494
687         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
688         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
689
690         PR middle-end/30421
691         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
692         If lower_lastprivate_clauses emits some statements, append them
693         to dlist rather than body_p and to body_p append an initializer.
694         (lower_omp_for): Adjust caller.
695
696 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
697
698         * target.h (globalize_decl_name): New.
699         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
700         * output.h (default_globalize_decl_name): New.
701         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
702         globalize_label.
703         (globalize_decl): Ditto.
704         (default_globalize_decl_name): New.
705         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
706         (ia64_handle_version_id_attribute): New.
707         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
708         (ia64_asm_output_external): Use globalize_decl_name instead
709         of globalize_label.
710         * doc/extend.texi (version_id): New pragma.
711         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
712
713 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
714
715         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
716         _uleb128_t and _Unwind_SWord with _sleb128_t.
717         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
718         uw_frame_state_for, uw_update_context_1): Likewise.
719         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
720         * unwind-pe.h (read_uleb128, read_sleb128,
721         read_encoded_value_with_base): Likewise.
722         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
723
724 2007-01-24  Richard Guenther  <rguenther@suse.de>
725
726         * builtins.c (expand_builtin_cexpi): Get the fndecl
727         for cexp in the correct way.
728
729 2007-01-24  Jan Hubicka  <jh@suse.cz>
730
731         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
732
733 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
734
735         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
736         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
737         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
738         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
739         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
740         __SIZEOF_PTRDIFF_T__.
741         * doc/cpp.texi: Documentation for the new builtin macros added.
742
743 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
744
745         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
746         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
747         instead of operands[2] to avoid access past the end of array.
748
749 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
750
751         * reload1.c (emit_reload_insns): Pass the reload register
752         for a non-spill output reload through forget_old_reloads_1.
753
754 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
755
756         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
757         _SOFT_DOUBLE if doubles use software floating-point.
758         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
759         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
760         * config/rs6000/darwin-ldouble.c: Also compile functions for
761         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
762         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
763         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
764         commented-out long double override.
765         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
766         * config/rs6000/eabispe.h: Likewise.
767         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
768         long double for non-SPE.
769         (rs6000_handle_option): Likewise.
770         (invalid_e500_subreg): Disallow more subregs involding DImode,
771         DFmode, TImode or TFmode.
772         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
773         E500 double.
774         (legitimate_lo_sum_address_p): Also check for TFmode for E500
775         double.
776         (rs6000_legitimize_address): Also handle TFmode for E500 double.
777         (rs6000_legitimize_reload_address): Also handle TFmode for E500
778         double.
779         (rs6000_legitimate_address): Also check for TFmode for E500
780         double.
781         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
782         (spe_build_register_parallel): Handle TFmode and TCmode.
783         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
784         double.
785         (function_arg): Handle TFmode and TCmode for E500 double.
786         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
787         in general.
788         (print_operand): Handle TFmode and TImode for %y.
789         (rs6000_generate_compare): Handle TFmode comparisons for E500
790         double.
791         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
792         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
793         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
794         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
795         TFmode for E500 double.
796         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
797         (floatsidf2): Enable for E500 double.
798         (movtf_softfloat): Use rs6000_nonimmediate_operand.
799         (extenddftf2): Change to extenddftf2_fprs.
800         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
801         depending on TARGET_E500_DOUBLE.
802         (extendsftf2): Enable for E500 double.
803         (trunctfdf2): Enable for E500 double.
804         (trunctfsf2): Change to trunctfsf2_fprs.
805         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
806         depending on TARGET_E500_DOUBLE.
807         (floatsitf2): Enable for E500 double.
808         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
809         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
810         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
811         (negtf2): Change to negtf2_internal.
812         (negtf2): New expander.
813         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
814         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
815         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
816         (movdi_internal32): Use rs6000_nonimmediate_operand.
817         (unnamed splitter): Likewise.
818         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
819         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
820         (SPE64TF, DITI): New mode macros.
821         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
822         modes.
823         (frob_tf_ti): New.
824         (frob_<mode>_di_2): New.
825         (frob_tf_di_8_2): New.
826         (frob_di_df): Change to frob_di_<mode>; allow more modes.
827         (frob_ti_tf): New.
828         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
829         more modes.
830         (frob_ti_<mode>_8_2): New.
831         (frob_ti_tf_2): New.
832         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
833         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
834         TFmode.
835         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
836         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
837         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
838         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
839         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
840         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
841         tsttflt_gp): New.
842
843 2007-01-23  Ian Lance Taylor  <iant@google.com>
844
845         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
846         insn-*.o first.
847         (OBJS-archive): Reformat, alphabetize.
848         (OBJS): Change out_object_file to OBJS-md.
849
850 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
851
852         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
853         JUMP_P and CALL_P predicates where applicable.
854         * config/i386/i386.c: Ditto.
855         * config/i386/i386.md: Ditto.
856         * config/i386/mmx.md: Ditto.
857         * config/i386/predicates.md: Ditto.
858
859 2007-01-22  Andreas Schwab  <schwab@suse.de>
860
861         * config/m68k/m68k.h: Fix comment.
862
863 2007-01-22  Jan Hubicka  <jh@suse.cz>
864
865         * passes.c (init_optimization_passes): Do not rerun
866         pass_early_warn_uninitialized.
867
868 2007-01-22  Richard Guenther  <rguenther@suse.de>
869
870         PR tree-optimization/30038
871         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
872         function.
873         (execute_cse_sincos_1): Likewise.
874         (execute_cse_sincos): Likewise.
875         (gate_cse_sincos): Likewise.
876         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
877         the cexpi() canonicalization of sincos().
878         * tree-pass.h (pass_cse_sincos): Declare.
879         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
880
881 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
882
883         PR rtl-optimization/29329
884         * combine.c (replaced_rhs_insn): Rename to i2mod.
885         (replaced_rhs_value): Rename to i2mod_new_rhs.
886         (i2mod_old_rhs): New global variable.
887         (combine_instructions): Adjust for above change.  Save a copy of
888         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
889         note are substituted in the second instruction.
890         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
891         ditch the note if it pertains to the second eliminated register
892         and this register is mentioned in i2mod_old_rhs.
893
894         Revert:
895         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
896
897         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
898         the insn to which the note was originally attached.
899
900 2007-01-21  Jan Hubicka  <jh@suse.cz>
901
902         * ipa-inline.c (inlining_mode): Comment, move up.
903         (cgraph_decide_inlining_incrementally): Do not perform inlining
904         itself; fix handling of flattening of self recursive functions.
905         (cgraph_find_cycles): Remove.
906         (cgraph_flatten_node): Remove.
907         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
908         (try_inline): New function.
909         (cgraph_early_inlining): Update call of
910         cgraph_decide_inlining_incrementally.  Apply inlining here.
911         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
912
913 2007-01-21  Dirk Mueller  <dmueller@suse.de>
914
915         PR bootstrap/30511
916         * tree-vrp.c (check_array_bounds): do not warn
917         about ADDR_EXPR's of ARRAY_REF's which are immediately
918         used in binary expressions.
919
920 2007-01-21  Ira Rosen  <irar@il.ibm.com>
921
922         * tree-vectorizer.h (struct _stmt_vec_info): Add new field
923         read_write_dep and macros for its access.
924         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
925         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
926         argument, call vect_check_interleaving for every independent pair of
927         data-refs. Mark loads that access the same memory location as a store
928         in the loop.
929         (vect_check_dependences): Remove.
930         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences
931         call, fix the call to vect_analyze_data_ref_dependence.
932         (vect_analyze_data_ref_access): For statements that access the same
933         data-ref, check that they are not stores; for loads, check that there
934         is no store that access the same location.
935
936 2007-01-20  Roger Sayle  <roger@eyesopen.com>
937             Joseph Myers  <joseph@codesourcery.com>
938
939         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
940         (-fdump-rtl-stack): Likewise.
941         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
942         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
943         (-mpt-fixed): Use @minus{} for minus sign.
944         (Using Precompiled Headers): Fix under/overfull hbox.
945
946 2007-01-20  Jan Hubicka  <jh@suse.cz>
947
948         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
949         place within annotation.
950
951 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
952
953         * pa.c (output_move_double): Change array size of xoperands to 4.
954
955 2007-01-20  Jan Hubicka  <jh@suse.cz>
956
957         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
958         as temporary.
959
960 2007-01-19  Ian Lance Taylor  <iant@google.com>
961
962         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
963         constant to unsigned HOST_WIDE_INT before negating.
964
965 2007-01-19  Ian Lance Taylor  <iant@google.com>
966
967         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
968         to unsigned.
969         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
970         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
971         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
972         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
973         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
974         (realloc_vdef, realloc_vuse): Change second parameter to
975         unsigned.
976         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
977         fields to unsigned.
978         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
979         unsigned. Change x and lim locals to unsigned.
980         (realloc_vdef, realloc_vuse): Change num_elem parameter to
981         unsigned.
982         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
983         (copy_virtual_operands): Change i and n locals to unsigned.
984
985 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
986
987         PR middle-end/29335
988         * builtins.c (fold_builtin_1): Handle builtin fdim.
989
990 2007-01-20  Jan Hubicka  <jh@suse.cz>
991
992         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
993         * tree-ssa-structalias.c (compute_points_to_sets): Do call
994         init_alias_heapvars.
995         (init_alias_heapvars): Initialize only when not already initialized.
996         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
997
998 2007-01-19  Roger Sayle  <roger@eyesopen.com>
999
1000         * common.opt (fsigned-zeros): New command line option.
1001         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
1002         of flag_unsafe_math_optimizations.
1003         * opts.c (set_fast_math_flags): The -ffast-math command line option
1004         implies -fno-signed-zeros.
1005         (fast_math_flags_set_p): Likewise.
1006
1007         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
1008         the documentation of -ffast-math appropriately.  Wrap long lines.
1009
1010 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
1011
1012         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
1013         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
1014
1015 2007-01-19  Tomas Bily  <tbily@suse.cz>
1016
1017         * cgraphunit.c (cgraph_finalize_function): Updating of pid
1018         * tree-profile.c:
1019         (tree_init_ic_make_global_vars): New function
1020         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
1021         (tree_gen_ic_profiler): New function
1022         (tree_gen_ic_func_profiler): New function
1023         (tree_profiling): Added calling of tree_gen_ic_func_profiler
1024         (tree_profile_hooks): Added hook for indirec/virtual calls
1025         * value-prof.c (tree_find_values_to_profile): New case for
1026         indirect calls
1027         (tree_values_to_profile): Call for determining indirect/virtual
1028         counters
1029         (tree_indirect_call_to_profile): New function
1030         (tree_ic_transform): New function
1031         (tree_ic): New function
1032         (find_func_by_pid): New function
1033         (init_pid_map): New function
1034         (tree_value_profile_transformations): Added check for
1035         indirect/virtual call transformation
1036         * value-prof.h (enum hist_type): New counter type for
1037         indirect/virtual calls
1038         (profile_hooks): Added new hook for profiling indirect/virtual
1039         calls
1040         * profile.c (instrument_values): New case for indirect/virtual
1041         call added
1042         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
1043         (GCOV_COUNTER_V_INDIR): New counter type
1044         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
1045         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
1046         call added
1047         * cgraph.c: Definition of cgraph_max_pid
1048         (cgraph_create_node): Default init of pid attribute
1049         * cgraph.h: Declaration of cgraph_max_pid
1050         (struct cgraph_node): Added pid attribute
1051         * libgcov.c (__gcov_indirect_call_profiler): New function
1052         (__gcov_one_value_profiler_body): New function
1053         (__gcov_one_value_profiler): Body was moved to
1054         __gcov_one_value_profiler_body and calls it
1055
1056 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
1057
1058         * doc/gty.texi (Options): Document the mark_hook option to GTY.
1059         * gengtype.c (write_types_data, write_func_for_structure,
1060         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
1061         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
1062         (walk_type, write_func_for_structure): Generate the mark_hook if
1063         needed.
1064
1065 2007-01-19  Jan Hubicka  <jh@suse.cz>
1066
1067         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1068         'early' argument take inlining mode argument specifying whether to
1069         inline for size/speeed or all functions; add support for flattening;
1070         improve dumpting.
1071         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1072
1073 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1074
1075         PR c++/17947
1076         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1077         the name of the declared identifier.
1078
1079 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1080
1081         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1082         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1083         duplicate evaluation.
1084
1085 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1086
1087         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1088         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
1089         to XFMode, use acosxf2 and truncate result to requested mode.
1090         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1091         (asin<mode>2): Similarly, with asin expanders.
1092         (*fscalexf4_i387): Rename from *fscalexf4.
1093         (expNcorexf3): New expander.
1094         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1095         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1096         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
1097         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1098         patterns for SSE math.
1099         (exp10<mode>2): Similarly, with exp10 expanders.
1100         (exp2<mode>2): Similarly, with exp2 expanders.
1101         (expm1<mode>2): Similarly, with expm1 expanders.
1102         (ldexp<mode>3): Similarly, with ldexp expanders.
1103         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1104         Use gen_truncxf<mode>2_i387_noop to truncate result.
1105
1106 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1107
1108         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1109         for ColdFire targets.
1110
1111 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
1112             Richard Sandiford  <richard@codesourcery.com>
1113
1114         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1115         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1116         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1117         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1118         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1119         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1120
1121 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1122
1123         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1124         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
1125         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
1126         (*adddi_dilshr32_cf): New pattern.
1127         (adddi3, subdi3): Remove first alternatives.
1128
1129 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1130
1131         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1132         implemented using move.l, do not use its cc result for floating-point
1133         comparisons.
1134
1135 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1136
1137         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1138         from accepting 's' constraints if flag_pic.
1139
1140 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1141
1142         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1143         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1144         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1145         condition from TARGET_68881 to TARGET_HARD_FLOAT.
1146
1147 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
1148
1149         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1150         ColdFire alternatives.
1151         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1152         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
1153         architectures that support ff1 instruction.
1154
1155 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1156             Julian Brown  <julian@codesourcery.com>
1157
1158         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1159         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1160         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
1161         Add commentary.
1162
1163 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
1164
1165         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1166         and return a bool.
1167         (output_move_const_into_data_reg, output_move_simode_const): Delete.
1168         * config/m68k/m68k.c (const_method, const_int_cost): Take a
1169         HOST_WIDE_INT instead of an rtx.
1170         (m68k_rtx_costs): Update call accordingly.
1171         (output_move_const_into_data_reg): Likewise.  Fix formatting.
1172         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1173         Return a bool.
1174         (output_move_simode_const): Update calls after above changes.
1175         Rework to use automatic variables and predicates like MEM_P.
1176         * config/m68k/m68k.md (pushexthisi_const): Update call to
1177         valid_mov3q_const.
1178
1179 2007-01-19  Dirk Mueller  <dmueller@suse.de>
1180
1181         * tree-ssa-alias.c (perform_var_substitution): Fix typo
1182         in dump_flags test.
1183
1184 2007-01-19  Richard Guenther  <rguenther@suse.de>
1185
1186         * builtins.c (expand_builtin_cexpi): Fall back to expanding
1187         via cexp in case sincos is not available.
1188
1189 2007-01-19  Richard Guenther  <rguenther@suse.de>
1190
1191         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1192         * defaults.h (TARGET_HAS_SINCOS): Default to off.
1193         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1194         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1195         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1196         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1197         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1198         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1199
1200 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1201
1202         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1203         New insn patterns.
1204         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1205         (atan2xf3): Directly generate RTL pattern.
1206         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1207         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1208         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1209         disable patterns for SSE math.
1210         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1211         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1212         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1213         disable patterns for SSE math.
1214
1215 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1216
1217         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1218
1219 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1220
1221         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1222         VIEW_CONVERT_EXPR to the same type as it's operand.
1223
1224 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1225
1226         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1227         configured for long double 128.
1228
1229 2007-01-18  Mike Stump  <mrs@apple.com>
1230
1231         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1232         error.
1233
1234 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1235
1236         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1237         type correct.
1238
1239 2007-01-18  Jan Hubicka  <jh@suse.cz>
1240
1241         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1242         greater than NUM_VOP_FREE_BUCKETS.
1243
1244 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1245
1246         * tree-ssa-structalias.c: Update comments.
1247         (ptabitmap_obstack): Removed.
1248         (pta_obstack): New.
1249         (oldpta_obstack): Ditto.
1250         (stats): Add a few members.
1251         (struct variable_info): Remove node, complex, address_taken, and
1252         indirect_target members. Add oldsolution member.
1253         (new_var_info): Do not initialize removed members.
1254         (constraint_expr_type): Remove INCLUDES.
1255         (constraint_graph): Add size, implicit_preds, rep,
1256         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1257         (FIRST_REF_NODE): New macro.
1258         (LAST_REF_NODE): Ditto.
1259         (FIRST_ADDR_NODE): Ditto.
1260         (find): New function.
1261         (unite): Ditto.
1262         (dump_constraint): Do not handle INCLUDES.
1263         (insert_into_complex): Do not insert duplicate constraints.
1264         (condense_varmap_nodes): Renamed and rewritten into ...
1265         (merge_node_constraints): This. Also fix bug in handling of
1266         offseted copy constraints.
1267         (clear_edges_for_node): No longer need to deal with preds at all,
1268         or removing associated preds/succs.
1269         (merge_graph_nodes): Deal with indirect_cycles.
1270         Don't deal with predecessors.
1271         (add_implicit_graph_edge): New function.
1272         (add_pred_graph_edge): Ditto.
1273         (add_graph_edge): Don't deal with predecessors.
1274         (build_constraint_graph): Removed.
1275         (build_pred_graph): New function.
1276         (build_succ_graph): Ditto.
1277         (struct scc_info): Removed in_component. Added roots, dfs, and
1278         node_mapping. Remove visited_index, unification_queue.
1279         (scc_visit): Deal with union-find we do now.
1280         Deal with cycles with REF nodes.
1281         (collapse_nodes): Renamed and rewritten to ...
1282         (unify_nodes): This.
1283         (process_unification_queue): Removed.
1284         (topo_visit): Cleanup
1285         (do_da_constraint): Use find.
1286         (do_sd_constraint): Ditto.
1287         (do_ds_constraint): Ditto.
1288         (do_complex_constraint): Ditto.
1289         (init_scc_info): Update for removed and added members.
1290         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1291         (find_indirect_cycles): This.
1292         (equivalence_class): New variable.
1293         (label_visit): New function.
1294         (perform_variable_substitution): Rewritten.
1295         (free_var_substitution_info): New function.
1296         (find_equivalent_node): Ditto.
1297         (move_complex_constraints): Ditto.
1298         (eliminate_indirect_cycles): Ditto.
1299         (solve_graph): Only propagate changed bits.
1300         Use indirect cycle elimination.
1301         Use find.
1302         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1303         (tree_id_eq): Renamed to ...
1304         (tree_vi_eq): This. Update for member change
1305         (insert_id_for_tree): Renamed and rewritten to ...
1306         (insert_vi_for_tree): This.
1307         (lookup_id_for_tree): Renamed and rewritten to ...
1308         (lookup_vi_for_tree): This.
1309         (get_id_for_tree): Renamed and rewritten to ...
1310         (get_vi_for_tree): Ditto.
1311         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1312         (process_constraint): Don't handle INCLUDES.
1313         Remove special ADDRESSOF case.
1314         (find_func_aliases): Rewrite to use vi functions instead of id
1315         ones.
1316         (create_function_info_for): Ditto.
1317         (create_variable_info_for): Ditto.
1318         (intra_create_variable_infos): Ditto.
1319         (merge_smts_into): Ditto.
1320         (find_what_p_points_to): Ditto.
1321         (init_base_vars): Ditto.
1322         (init_alias_vars): Ditto.
1323         (remove_preds_and_fake_succs): New function.
1324         (dump_sa_points_to_info): Dump new stats.
1325         (dump_solution_for_var): Use find.
1326         (set_used_smts): Fix formatting.
1327         (compute_points_to_sets): Updated for new functions.
1328         (ipa_pta_execute): Ditto.
1329
1330 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1331             Richard Sandiford  <richard@codesourcery.com>
1332
1333         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1334         parallels.
1335         * calls.c (expand_call): If the return value is a PARALLEL,
1336         extract its first member.
1337         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1338         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1339         use FUNCTION_EXTRA_EPILOGUE.
1340         (m68k_function_value): Return a PARALLEL if the return value
1341         is of a pointer type.
1342         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1343         (FUNCTION_EXTRA_EPILOGUE): Remove.
1344         * config/m68k/m68k.md (D0_REG): New constant.
1345
1346 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1347
1348         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1349         output a NOP for empty epilogues.
1350
1351 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1352
1353         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1354         before function.  Extend register save check to include all
1355         registers, not just integer ones.
1356
1357 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1358
1359         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1360         (m68k_use_return_insn): ...this.
1361         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1362         * config/m68k/m68k.c (use_return_insn): Rename to...
1363         (m68k_use_return_insn): ...this.
1364         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1365         of USE_RETURN_INSN.
1366
1367 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1368
1369         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1370         nearest even, fix denormal rounding overflow.
1371
1372 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1373
1374         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1375         f%$smove and f%$move.
1376         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1377         (extendsfdf2_cf): Use fdmove instead of f%&move.
1378         (truncdfsf2_cf): Use fsmove instead of f%$smove.
1379         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1380
1381 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1382
1383         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1384         GPR<-GPR moves.
1385
1386 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1387
1388         * real.h (real_format): Add a canonical_nan_lsbs_set field.
1389         (coldfire_single_format): Declare.
1390         (coldfire_double_format): Likewise.
1391         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1392         of qnan_msb_set to determine the lower bits of a canonical
1393         NaN significand.
1394         (encode_ieee_double): Likewise.
1395         (encode_ieee_quad): Likewise.
1396         (ieee_single_format): Initialize canonical_nan_lsbs_set.
1397         (mips_single_format): Likewise.
1398         (ieee_double_format): Likewise.
1399         (mips_double_format): Likewise.
1400         (ieee_extended_motorola_format): Likewise.
1401         (ieee_extended_intel_96_format): Likewise.
1402         (ieee_extended_intel_128_format): Likewise.
1403         (ieee_extended_intel_96_round_53_format): Likewise.
1404         (ibm_extended_format): Likewise.
1405         (mips_extended_format): Likewise.
1406         (ieee_quad_format): Likewise.
1407         (mips_quad_format): Likewise.
1408         (vax_f_format): Likewise.
1409         (vax_d_format): Likewise.
1410         (vax_g_format): Likewise.
1411         (i370_single_format): Likewise.
1412         (i370_double_format): Likewise.
1413         (decimal_single_format): Likewise.
1414         (decimal_double_format): Likewise.
1415         (decimal_quad_format): Likewise.
1416         (c4x_single_format): Likewise.
1417         (c4x_extended_format): Likewise.
1418         (real_internal_format): Likewise.
1419         (coldfire_single_format): New real_format.
1420         (coldfire_double_format): Likewise.
1421         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1422         canonical_nan_lsbs_set.
1423         (pdp11_d_format): Likewise.
1424         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1425         if TARGET_COLDFIRE_CPU.
1426
1427 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1428
1429         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1430         (output_call, m68k_legitimize_call_address): Declare.
1431         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1432         parenthesees.  Add support for a 'W' constraint.
1433         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1434         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1435         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1436         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1437         variables.
1438         (override_options): Initialize them.  Do not set flag_no_function_cse
1439         for TARGET_ID_SHARED_LIBRARY.
1440         (m68k_output_pic_call): Delete.
1441         (m68k_legitimize_call_address): New function.
1442         (print_operand): Remove the %o prefix.  Handle the %p prefix.
1443         (output_call): New function.
1444         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
1445         target address from the GOT if symbolic jumps are not allowed.
1446         * config/m68k/m68k.md (call, general_operand): Do not set
1447         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
1448         Merge separate flag_pic and !flag_pic define_insns into...
1449         (*call, *call_value): ...these new patterns.  Match the address
1450         rather than the containing MEM and require it to be a call_operand.
1451         Use output_call to generate the asm template.
1452         * config/m68k/predicates.md (const_call_operand): New predicate.
1453         (call_operand): Likewise.
1454
1455 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1456
1457         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1458         (STACK_GROWS_DOWNWARD): Define to 1.
1459         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1460         comments, emphasizing that these values are only defaults.
1461         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1462         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1463
1464 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1465
1466         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1467
1468 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1469             Richard Sandiford  <richard@codesourcery.com>
1470
1471         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1472         (FP_REGNO_P): New macros.
1473         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1474         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1475         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1476         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1477
1478 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1479
1480         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1481         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1482         to $tmake_file.
1483         * config/m68k/t-floatlib: New file, extracting common code from...
1484         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1485         * config/m68k/t-uclinux: Here.
1486         * config/m68k/fpgnulib.c: Do not compile extendeed precision
1487         routines on ColdFire targets.
1488
1489 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1490
1491         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1492         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1493         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1494         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1495         explicitly set MOTOROLA to 1.
1496         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1497
1498 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1499             Nathan Sidwell  <nathan@codesourcery.com>
1500
1501         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1502         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1503         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1504         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1505         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1506         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1507         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1508         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1509         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1510         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1511         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1512         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1513         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1514         (DBX_REGISTER_NUMBER): Delete.
1515         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1516         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1517         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1518
1519 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1520
1521         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1522         * config/flat.h: New file.
1523         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1524         OBJECT_FORMAT_FLAT.
1525         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1526         indirect references for -msep-data or -mid-shared-library.
1527         Do not use PC-relative code addresses either.
1528         * config/m68k/m68k.c (override_options): Restrict -fPIC error
1529         to -mpcrel.
1530         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
1531         for shared libraries and crt1.o for executables.  Use crti.o and
1532         crtbegin.o.
1533         (ENDFILE_SPEC): Use crtend.o and crtn.o.
1534         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1535         Do not add -elf2flt or -shared-lib-id options here.
1536         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
1537         Pass -shared-lib-id if -mid-shared-library, taking the library
1538         identifier from -mshared-library-id if given, otherwise
1539         defaulting to 0.
1540         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1541         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1542         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1543         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1544         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1545         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1546         and crtend.o.
1547         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1548         jump sequence for ISA A and ISA A+.
1549         (PICJUMP): Likewise.
1550
1551 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1552
1553         * config.gcc (m68k-*-uclinux*): Base the port on the common
1554         and m68k GNU/Linux files rather than on the generic ELF ones.
1555         * config/m68k/uclinux.h (TARGET_VERSION): Override.
1556         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1557
1558 2007-01-18  Julian Brown  <julian@codesourcery.com>
1559             Richard Sandiford  <richard@codesourcery.com>
1560
1561         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1562         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1563         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1564         before redefining.
1565         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1566         80 unconditionally.
1567         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1568
1569 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1570
1571         * doc/install.texi: Document m68k-uclinuxoldabi.
1572         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1573         * config/m68k/uclinux-oldabi.h: New file, copied from
1574         config/m68k/uclinux.h.
1575
1576 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
1577
1578         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1579         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1580         (m32r_compute_frame_size): Use unsigned for regno.
1581         (m32r_reload_lr): Use gen_frame_mem.
1582         (pop): New.
1583         (m32r_output_function_epilogue): Don't output the function epilogue
1584         textually here.
1585         (m32r_expand_epilogue): New.
1586         (direct_return): Return false if the function has the interrupt
1587         attribute.
1588         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1589         * config/m32r/m32r.md (epilogue): New expander.
1590         (return_lr, return_rte): New insns.
1591         (return): Make it expander.
1592         (return_normal): New expander.
1593
1594 2007-01-18  Josh Conner  <jconner@apple.com>
1595
1596         PR target/30485
1597         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1598         support for UNLE, UNLT, UNGE, and UNGT.
1599
1600 2007-01-18  Jan Hubicka  <jh@suse.cz>
1601
1602         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1603         * tree-tailcall (add_virtual_phis): Likewise.
1604         (optimize_tail_call): Return TODOs.
1605         (execute_tail_calls): Return TODOs.
1606         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1607         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1608         changed.
1609         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1610         Cleanup cfg using TODO.
1611         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1612         * passes.c (execute_function_todo): When cleanup did something, remove
1613         unused locals.
1614         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1615         (make_edges): Don't cleanup_cfg.
1616
1617 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
1618
1619         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1620         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1621         operate on the top of stack.
1622         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1623         register handling of instructions that output to the second
1624         stack slot.
1625         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1626         (move_for_stack_reg): Special-case check for dead destination
1627         stack slot for constant load of 1.0 inside UNSPEC_TAN.
1628
1629         * config/i386/i386.md (UNSPEC_TAN): New constant.
1630         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1631         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1632         to correctly model move of constant 1.0 to top stack slot.
1633         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1634         (unnamed peephole2 pattern): Remove corresponding peephole2
1635         pattern that optimizes tan insn and loading of constant 1.0.
1636         (tanxf2): Use fptanxf4_i387.
1637         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1638         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
1639         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1640         disable patterns for SSE math.
1641         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1642         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1643
1644 2007-01-18  Dirk Mueller  <dmueller@suse.de>
1645             Richard Guenther <rguenther@suse.de>
1646
1647         PR diagnostic/8268
1648         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1649         * common.opt (Warray-bounds): Add new warning option.
1650         * c-opts.c (c_common_handle_option): Define -Warray-bounds
1651         if -Wall is given.
1652         * Makefile.in: make tree-vrp.o depend on toplev.h
1653         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1654         is enabled.
1655         (check_array_refs, check_array_bounds, check_array_ref): New.
1656
1657 2007-01-18  Jan Hubicka  <jh@suse.cz>
1658
1659         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1660         (execute_ssa_ccp): Return flags conditionally.
1661         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1662         changed.
1663         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1664
1665 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
1666
1667         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1668         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1669
1670         * Makefile.c (GTFILES): Add cfglayout.h.
1671         * gengtype.c (open_base_files): Likewise.
1672         * cfglayout.c (cfg_layout_function_footer,
1673         cfg_layout_function_header) Reindent to make gengtype happy.
1674         * cfglayout.h (cfg_layout_function_footer,
1675         cfg_layout_function_header): Add GTY(()) marker.
1676
1677         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1678
1679 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1680
1681         * genautomata.c (write_automata): Include xstrerror output in the
1682         error message if writing the DFA description file fails.
1683
1684 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1685
1686         * config/mips/mips-protos.h (mips_output_external): Make it
1687         return void.
1688         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1689         * config/mips/mips.c (irix_output_external_libcall): Likewise.
1690         (extern_list): Likewise.
1691         (extern_head): Likewise.
1692         (TARGET_ASM_FILE_END): Likewise.
1693         (mips_file_end): Likewise.
1694         (mips_output_external): Rewritten.
1695
1696 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1697
1698         * genpreds.c (write_insn_preds_c): Only write out the function
1699         body for regclass_for_constraint if we have register constraints.
1700
1701 2007-01-17  Tom Tromey  <tromey@redhat.com>
1702
1703         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1704         * doc/install.texi (Testing): Use sourceware.org.
1705         (Binaries): Likewise.
1706         (Specific): Likewise.
1707         * doc/contrib.texi (Contributors): Use sourceware.org.
1708
1709 2007-01-17  Anatoly Sokolov <aesok@post.ru>
1710
1711         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1712         (AVR_ENHANCED): Rename to ...
1713         (AVR_HAVE_MUL): ... new.
1714         (avr_enhanced_p): Rename to ...
1715         (avr_have_mul_p): ... new.
1716         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1717         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1718         * config/avr/avr.c (avr_enhanced_p): Rename to ...
1719         (avr_have_mul_p): ... new.
1720         (base_arch_s): Rename 'enhanced' to 'have_mul'.
1721         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1722         'avr_enhanced_p' and 'enhanced'.
1723         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1724         AVR_HAVE_MUL instead of AVR_ENHANCED.
1725         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1726         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1727         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1728         instead of AVR_ENHANCED.
1729         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1730         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1731         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1732
1733 2007-01-17  Ian Lance Taylor  <iant@google.com>
1734
1735         * vec.h (VEC_reserve_exact): Define.
1736         (vec_gc_p_reserve_exact): Declare.
1737         (vec_gc_o_reserve_exact): Declare.
1738         (vec_heap_p_reserve_exact): Declare.
1739         (vec_heap_o_reserve_exact): Declare.
1740         (VEC_OP (T,A,reserve_exact)): New static inline function, three
1741         versions.
1742         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1743         negative parameter.
1744         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1745         (VEC_OP (T,A,copy)) [all versions]: Likewise.
1746         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1747         * vec.c (calculate_allocation): Add exact parameter.  Change all
1748         callers.
1749         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1750         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1751         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1752         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1753         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1754         (vec_heap_p_reserve_exact): New function.
1755         (vec_heap_o_reserve_exact): New function.
1756
1757 2007-01-17  Jan Hubicka  <jh@suse.cz>
1758
1759         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1760
1761 2007-01-17  Eric Christopher  <echristo@apple.com>
1762
1763         * config.gcc: Support core2 processor.
1764
1765 2007-01-16  Jan Hubicka  <jh@suse.cz>
1766
1767         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1768         the names, just unlink the chain so we don't crash on dangling pointers
1769         to dead SSA names.
1770
1771 2007-01-16  Jan Hubicka  <jh@suse.cz>
1772
1773         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1774         * tree-pass.h: Reorder to make IPA passes appear toegher.
1775         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1776         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1777         parameters, do not call early inliner.
1778         * ipa-inline.c: Update comments.  Include tree-flow.h
1779         (cgraph_decide_inlining): Do not compute inlining parameters.
1780         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1781         be called with function context set up.
1782         (pass_ipa_inline): Remove unreachable functions before pass.
1783         (cgraph_early_inlining): Simplify assuming to be called from the
1784         PM as local pass.
1785         (pass_early_inline): New pass.
1786         (cgraph_gate_ipa_early_inlining): New gate.
1787         (pass_ipa_early_inline): Turn into simple wrapper.
1788         (compute_inline_parameters): New function.
1789         (gate_inline_passes): New gate.
1790         (pass_inline_parameters): New pass.
1791         (apply_inline): Move here from tree-optimize.c
1792         (pass_apply_inline): New pass.
1793         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1794         transforming.
1795         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1796         doing them by hand.
1797         (tree_function_versioning): Do not allocate dummy struct function.
1798         * tree-inline.h (optimize_inline_calls): Update prototype.
1799         * tree-optimize.c (execute_fixup_cfg): Export.
1800         (pass_fixup_cfg): Remove
1801         (tree_rest_of_compilation): Do not apply inlines.
1802         * tree-flow.h (execute_fixup_cfg): Declare.
1803         * Makefile.in (gt-passes.c): New.
1804         * passes.c: Include gt-passes.h
1805         (init_optimization_passes): New passes.
1806         (nnodes, order): New static vars.
1807         (do_per_function_toporder): New function.
1808         (execute_one_pass): Dump current pass here.
1809         (execute_ipa_pass_list): Don't dump current pass here.
1810
1811 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
1812
1813         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1814         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1815
1816 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
1817
1818         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1819         (strong_alias): Define.
1820         (__gcc_qmul): Provide non-FMA for soft-float.
1821         (__gcc_qdiv): Same.
1822         (__gcc_qneg): New.
1823         (__gcc_qeq): New.
1824         (__gcc_qle): New.
1825         (__gcc_qge): New.
1826         (__gcc_qunord): New.
1827         (__gcc_stoq): New.
1828         (__gcc_dtoq): New.
1829         (__gcc_qtos): New.
1830         (__gcc_qtod): New.
1831         (__gcc_qtoi): New.
1832         (__gcc_qtou): New.
1833         (__gcc_itoq): New.
1834         (__gcc_utoq): New.
1835         (fmsub): New.
1836         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1837         soft-float functions.
1838         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1839         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1840         about long double soft float.
1841
1842 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
1843             Tehila Meyzels  <tehila@il.ibm.com>
1844
1845         * tree-vectorizer.h (is_pattern_stmt_p): New.
1846         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1847         formatting (tabs instead of spaces). Cleanup and clarify setting
1848         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1849         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1850         (vectorizable_type_demotion): Check that types are integral.
1851         (vectorizable_type_promotion): Likewise.
1852         (vectorizable_store): Fix typo.  Eliminate new-line at end of
1853         comments.
1854
1855 2007-01-16  Jan Hubicka  <jh@suse.cz>
1856
1857         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1858         cgraph edges too.
1859
1860 2007-01-15  Eric Christopher  <echristo@apple.com>
1861
1862         * ifcvt.c: Include vec.h, vecprim.h.
1863         (check_cond_move_block): New argument regs.
1864         Reorganize. Add registers used to regs.
1865         (cond_move_process_if_block): Use regs set above as
1866         loop bounds.
1867
1868 2007-01-15  Eric Christopher  <echristo@apple.com>
1869
1870         * config/darwin.h: Update copyright.
1871         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1872         (LINK_COMMAND_SPEC): Add -u.
1873         (LINK_SPEC): Fix umbrella for above.
1874
1875 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
1876
1877         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1878         glibc CVS.
1879
1880 2007-01-15  Tom Tromey  <tromey@redhat.com>
1881
1882         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1883         * doc/install.texi (Testing): Don't mention jacks.
1884         (Configuration): Document --enable-java-maintainer-mode.  Move
1885         --with-java-home to libgcj-specific section.  Document
1886         --with-ecj-jar.
1887         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1888
1889 2007-01-15  Jan Hubicka  <jh@suse.cz>
1890
1891         * tree-ssa-dce.c (DCE_TODOs): New.
1892         (propagate_necessity): Return if something changed.
1893         (eliminate_unnecessary_stmts): Likewise.
1894         (perform_tree_ssa_dce): Return TODO flags when needed.
1895         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1896
1897 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
1898
1899         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1900         (fyl2x_extend<mode>xf3_i387): New insn pattern.
1901         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1902         patterns using X87MODEF12 mode macro.  Extend operand 1
1903         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1904         (log10<mode>2): Ditto.
1905         (log2<mode>2): Ditto.
1906         (log1p<mode>2): Ditto.
1907         (logb<mode>2): Ditto.
1908         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1909         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1910         (*fxtractxf3_i387): Rename from *fxtractxf3.
1911         (fxtract_extend<mode>xf3_i387): New insn pattern.
1912         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1913         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1914         and gen_fyl2xxf3_i387().
1915
1916 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
1917
1918         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1919         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1920         get_loop_exit_edges.  Changed return type to bool.
1921         * cfgloop.h (fix_loop_placement): Declaration removed.
1922
1923 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1924
1925         * param.h (MIN_VECT_LOOP_BOUND): New.
1926         * params.def (MIN_VECT_LOOP_BOUND): New.
1927         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1928         argument - minimum threshold for number of iterations.
1929         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1930         argument to declaration.
1931         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1932         MIN_VECT_LOOP_BOUND.
1933         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1934         slpeel_tree_peel_loop_to_edge with additional argument.
1935         (vect_do_peeling_for_alignment): Likewise.
1936         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1937
1938 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1939
1940         PR target/30413
1941         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1942         operands of size 1.
1943
1944 2007-01-14  Jan Hubicka  <jh@suse.cz>
1945
1946         * tree-dfa.c (remove_referenced_var): New function.
1947         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1948         prune referenced vars list too.
1949         * tree-flow.h (remove_referenced_var): Declare.
1950
1951 2007-01-14  Jan Hubicka  <jh@suse.cz>
1952
1953         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1954         separately.
1955         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1956         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1957         * except.h (verify_eh_throw_table_statements): Kill prototype.
1958         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1959         into call exrepssions.
1960         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1961         eh trhow verifier.
1962         * tree-cfg.c: Include pointer-set.h.
1963         (verify_node_sharing): Work on pointer set.
1964         (verify_eh_throw_stmt_node): New.
1965         (verify_stmts): Use pointers sets, verify throw_stmt.
1966
1967 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1968
1969         * ipa-reference.c (analyze_function): Consider also addresses taken
1970         in phi nodes.
1971
1972 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1973
1974         * c-typeck.c (null_pointer_constant_p): Replace use of
1975         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1976         (build_c_cast): Likewise.
1977
1978 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1979
1980         * tree.h (force_fit_type_double): Remove unused final argument.
1981         * c-common.c (constant_expression_warning): Replace use of
1982         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1983         (convert_and_check): Likewise.
1984         (shorten_compare): Update call to force_fit_type_double.
1985         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1986         * convert.c (convert_to_pointer): Update call to
1987         force_fit_type_double.
1988         * fold-const.c (force_fit_type_double): Remove overflowed_const
1989         argument.
1990         (int_const_binop, fold_convert_const_int_from_int,
1991         fold_convert_const_int_from_real, fold_div_compare,
1992         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1993         fold_abs_const, fold_not_const): Remove the final argument from
1994         calls to force_fit_type_double.
1995
1996 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1997
1998         * configure.ac: Set insn to "nop" for spu-*-* also.
1999         * configure: Regenerate.
2000
2001 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
2002
2003         * builtins.def : Add BUILT_IN_FREE.
2004
2005 2007-01-12  Jan Hubicka  <jh@suse.cz>
2006
2007         PR tree-optimization/30443
2008         * tree-inline.c (tree_function_versioning): Do not optimize when
2009         cloning for inlining.
2010
2011 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
2012
2013         * doc/loop.texi: Document recording of loop exits.
2014         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
2015         (update_single_exits_after_duplication,
2016         update_single_exit_for_duplicated_loop,
2017         update_single_exit_for_duplicated_loops): Removed.
2018         (duplicate_loop_to_header_edge): Do not call
2019         update_single_exits_after_duplication and
2020         update_single_exit_for_duplicated_loops.
2021         (loop_version): Do not update single_exit information.
2022         (fix_loop_structure): Use record_loop_exits instead of
2023         mark_single_exit_loops.
2024         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
2025         the lists of loop exits.
2026         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
2027         split_edge, merge_blocks): Update the lists of loop exits.
2028         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
2029         loop_optimizer_init.
2030         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
2031         of mark_single_exit_loops.
2032         (loop_optimizer_finalize): Call release_recorded_exits.
2033         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
2034         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
2035         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
2036         update single exit information.
2037         * lambda-code.c (perfect_nestify): Ditto.
2038         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
2039         (mark_single_exit_loops): Removed.
2040         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
2041         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
2042         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
2043         functions.
2044         (get_loop_exit_edges, single_exit): Use recorded exit lists.
2045         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
2046         (verify_loop_structure): Verify consistency of the exit lists.
2047         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
2048         (set_single_exit): Removed.
2049         * cfgloop.h (struct loop_exit): New function.
2050         (struct loop): single_exit_ field replaced by exits field.
2051         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
2052         (struct loops): Added exits hash.
2053         (mark_single_exit_loops, set_single_exit): Declaration removed.
2054         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
2055
2056 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2057
2058         * doc/invoke.texi: Avoid use of @headitem.
2059
2060 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2061
2062         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
2063
2064 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2065
2066         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2067         list of targets.
2068
2069 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2070             Richard Sandiford  <richard@codesourcery.com>
2071
2072         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2073         * config/m68k/m68k.opt: Resort options.
2074         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2075
2076 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2077             Richard Sandiford  <richard@codesourcery.com>
2078
2079         * doc/invoke.texi: Document the macros that are defined by
2080         m68k's -mtune and -mhard-float options.
2081         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2082         (m68k_cpp_cpu_family): Likewise.
2083         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2084         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
2085         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
2086         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2087         (m68k_cpp_cpu_family): Likewise.
2088
2089 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2090
2091         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2092         macros besides mc68000 as tuning macros.  Use a switch statement
2093         to set them and mcpu32.
2094
2095 2007-01-12  Julian Brown  <julian@codesourcery.com>
2096
2097         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2098         (TARGET_68040_ONLY): Rename to...
2099         (TARGET_68040): ...this.
2100         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2101         * config/m68k/m68k.md: Likewise.
2102
2103 2007-01-12  Julian Brown  <julian@codesourcery.com>
2104             Nathan Sidwell  <nathan@codesourcery.com>
2105             Richard Sandiford  <richard@codesourcery.com>
2106
2107         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2108         the -mcpu= argument associated with the --with-cpu setting.
2109         Define M68K_DEFAULT_TUNE to the default -mtune= option,
2110         if different from the one implied by the -mcpu setting.
2111         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2112         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
2113         * doc/invoke.texi: Mention ColdFire in the introduction to the
2114         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
2115         -mno-div and -mhard-float options.  Make -m68881 a synonym for
2116         -mhard-float.  Document the previously-undocumented -m5206e,
2117         -m528x, -m5307 and -m5407 options.  Tweak the existing option
2118         documentation for consistency.
2119         * doc/install.texi: Mention new --with-cpu arguments.
2120         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2121         default CPU if neither -mcpu nor -march are specified.
2122         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2123         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2124         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
2125         using m68k_tune to decide between families that implement the
2126         same ISA.  Use m68k_tune to set __mcfv4e__.
2127         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2128         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2129         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2130         (FL_ISA_C, FL_ISA_MMU): New macros.
2131         (MASK_COLDFIRE): Delete.
2132         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2133         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2134         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2135         (TARGET_HARD_FLOAT): Do not define here.
2136         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2137         (TUNE_68000): New macro.
2138         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2139         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2140         (TUNE_CFV2): Redefine in terms of m68k_tune.
2141         (uarch_type, target_device, fpu_type): New enums.
2142         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2143         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2144         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2145         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2146         (FL_FOR_isa_c): New macros.
2147         (m68k_isa): New enum.
2148         (m68k_target_selection): New structure.
2149         (all_devices, all_isas, all_microarchs): New tables.
2150         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2151         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2152         (MASK_ALL_CPU_BITS): Delete.
2153         (m68k_find_selection): New function.
2154         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2155         Map the legacy target options to a combination of the new ones.
2156         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2157         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
2158         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2159         settings.
2160         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2161         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2162         properties.
2163         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2164         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2165         * config/m68k/m68k-devices.def: New file.
2166
2167 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2168             Nathan Sidwell  <nathan@codesourcery.com>
2169
2170         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2171         (SUBTARGET_EXTRA_SPECS): New macros.
2172         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2173         use %(asm_cpu_spec) instead.
2174         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2175         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2176         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2177         (EXTRA_SPECS): Rename to...
2178         (SUBTARGET_EXTRA_SPECS): ...this.
2179
2180 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
2181             Richard Sandiford  <richard@codesourcery.com>
2182             Julian Brown  <julian@codesourcery.com>
2183
2184         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2185         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2186         to the configuration's default CPU.
2187         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2188         Remove default masks.
2189         (m680[012]0-*-*): Set the default with_cpu to the first part of
2190         the target name.
2191         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2192         (m68k*-*-linux): Extend the --with-cpu handling to...
2193         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
2194         and m68010.  Don't set target_cpu_default2.
2195         * doc/install.texi: Document --with-cpu for m68k.
2196         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2197         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2198         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2199         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2200         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2201         (ASM_SPEC): Remove use of %(asm_cpu_default).
2202         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2203         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2204         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2205         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2206         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2207         (TARGET_DEFAULT): Delete.
2208         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2209         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2210         (CPP_CPU_SPEC): Delete.
2211         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2212         the appropriate string.
2213         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2214         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2215         (ASM_SPEC): Don't use %(asm_default_spec).
2216         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2217         TARGET_DEFAULT and add MASK_68881.
2218         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2219
2220 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2221
2222         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2223         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2224         MASK_68010 alongside MASK_68020.
2225         * doc/invoke.texi: Document -m68010.
2226         * config/m68k/m68k.opt (m68010): New.
2227         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2228         if TUNE_68010.
2229         (TUNE_68010): New macro.
2230         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2231         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2232         (M68K_CPU_m68332): Add MASK_68010.
2233         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2234         fallback definition.
2235         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2236         defines.
2237         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2238         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2239         to all entries that use MASK_68020.
2240         (output_move_simode_const, output_move_himode, output_move_qimode)
2241         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2242         instead of TARGET_68020 to select clr behavior.  Remove comment
2243         about there being no TARGET_68010.
2244         * config/m68k/m68k.md: Likewise throughout.
2245
2246 2007-01-12  Julian Brown  <julian@codesourcery.com>
2247
2248         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2249         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2250         * config/m68k/m68k.md: Likewise.
2251
2252 2007-01-12  Julian Brown  <julian@codesourcery.com>
2253
2254         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2255         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2256
2257 2007-01-12  Julian Brown  <julian@codesourcery.com>
2258
2259         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2260         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2261         * config/m68k/m68k.md: Likewise.
2262
2263 2007-01-12  Julian Brown  <julian@codesourcery.com>
2264             Richard Sandiford  <richard@codesourcery.com>
2265
2266         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2267         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2268         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2269         TARGET_CPU32.
2270         (TARGET_CPU32): Rename to...
2271         (TUNE_CPU32): ...this.
2272         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2273         (TUNE_CFV2): New macros.
2274         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2275         remove conditions that are implied by TARGET_68020.
2276         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2277         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2278         (m68k_output_function_epilogue): Likewise.
2279         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2280         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2281         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2282         non-68000 timings.  Refactor multiplication and division costs.
2283         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2284         TUNE_CPU32 instead of TARGET_CPU32.
2285         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2286         and TUNE_68060 instead of TARGET_68060.
2287         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2288         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2289         TARGET_CPU32.
2290         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2291         && !TARGET_COLDFIRE" to choose between moveq and clr.
2292         Likewise in the unnamed movsf pattern.
2293         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2294         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2295         ashiftrt pattern.
2296
2297 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2298
2299         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2300         of tabbing before backslashes.
2301
2302 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2303
2304         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2305         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2306         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2307
2308 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2309
2310         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2311         find_interesting_uses_cond.
2312         (find_interesting_uses_cond): Use extract_cond_operands.
2313         (rewrite_use_compare): Use extract_cond_operands and
2314         force_gimple_operand_bsi.  Do not call update_stmt.
2315         (determine_use_iv_cost_condition): Use extract_cond_operands.
2316         Return cheaper of using original bound and changing the exit bound.
2317
2318 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2319
2320         PR tree-optimization/29516
2321         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2322         most_expensive_mult_to_index, addr_to_parts,
2323         create_mem_ref, maybe_fold_tmr): Make the type of
2324         fields of TARGET_MEM_REF sizetype.
2325         (move_fixed_address_to_symbol, move_pointer_to_base):
2326         New functions.
2327         * tree.def (TARGET_MEM_REF): Add comment on types of
2328         the operands.
2329
2330 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
2331
2332         * c-common.c (vector_types_convertible_p): Treat opaque types as
2333         always convertible if they have the same size, but not otherwise.
2334
2335 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
2336
2337         * ifcvt.c (struct noce_if_info): Add comments to the fields.
2338         Remove the b_unconditional field.
2339         (noce_try_sign_mask): Do not look at b_unconditional.
2340         (noce_process_if_block): Do not use merge_if_blocks.  Update
2341         the CFG here.  Do not set b_unconditional.
2342         (cond_move_process_if_block): Likewise.
2343         (find_cond_trap): Likewise.
2344         (check_cond_move_block): Require simple jump insns at the end
2345         of the basic block.
2346
2347 2007-01-11  Jan Hubicka  <jh@suse.cz>
2348
2349         PR tree-optimization/1046
2350         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2351         when alias info is not ready.
2352         (pass_tail_recursion): Do not require aliasing.
2353         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2354         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2355         * tree-ssa-copy.c (pass_copy_prop): Likewise.
2356         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2357         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2358         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2359         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2360         before inlining.
2361         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2362         When aliasing is not build, mark statement as volatile.
2363
2364 2007-01-11  Tom Tromey  <tromey@redhat.com>
2365
2366         PR preprocessor/15185, PR preprocessor/20989:
2367         * doc/cppopts.texi <-MT>: Update description of algorithm for
2368         computing default target.
2369         <-M, -MD>: Reword "basename" text.
2370
2371 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2372
2373         * builtins.c (expand_builtin_pow, expand_builtin_powi,
2374         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2375         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2376         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2377         fold_builtin_bswap, real_constp, fold_builtin_pow,
2378         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2379         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2380         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2381         * convert.c (convert_to_pointer): Likewise.
2382         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2383         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2384         const_binop, fold_convert_const_int_from_int,
2385         fold_convert_const_int_from_real,
2386         fold_convert_const_real_from_real, sign_bit_p,
2387         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2388         fold_sign_changed_comparison, fold_unary, fold_comparison,
2389         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2390         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2391         * print-tree.c (print_node_brief, print_node): Likewise.
2392         * stor-layout.c (place_field, layout_type): Likewise.
2393         * tree-chrec.c (keep_cast): Likewise.
2394         * tree.c (build_vector, build_real, build_real_from_int_cst,
2395         build_complex): Likewise.
2396
2397 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2398
2399         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
2400         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2401
2402 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
2403
2404         * configure.ac (strict1_warn): Rename to strict_warn.
2405         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2406         cc_set_by_configure, quoted_cc_set_by_configure,
2407         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2408         all_boot_languages, all_stagestuff): Remove.
2409         (target_list): Remove bootstrap targets.
2410         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2411         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2412         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2413         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2414         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2415         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2416         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2417         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2418         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2419         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2420         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2421         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2422         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2423         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2424         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2425         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2426         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2427         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2428         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2429         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2430         (mostlyclean): Adjust.
2431         (clean, distclean): Don't mention bootstrap stuff.
2432         * configure: Regenerate.
2433         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2434         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2435         treelang/config-lang.in (stagestuff): Remove.
2436         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2437         stageprofile, stagefeedback, stagestuff): Remove mention.
2438
2439 2007-01-11  Nick Clifton  <nickc@redhat.com>
2440
2441         * config/mcore/predicates.md (mcore_general_movesrc_operand):
2442         Accept CONSTs.
2443         (mcore_general_movdst_operand): Do not accept CONST_INTs.
2444         (mcore_arith_K_S_operand): Run the test for the S constraint not
2445         the test for the M constraint.
2446         (mcore_addsub_operand): Do not accept integer values that are
2447         larger than 32 bits.
2448         * config/mcore/mcore.md: Remove unused constraints from split.
2449         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2450         (addsi3): Likewise.
2451         (allocate_stack): Likewise.
2452         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2453         operands to 32 bits.
2454         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2455         INTVAL.
2456         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2457         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2458         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2459         mcore_output_bclri, mcore_output_andn, output_inline_const,
2460         mcore_output_move, mcore_output_movedouble): Likewise.
2461         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2462         (output_inline_const): Likewise.
2463         (output_inline_const): Fix format strings used in sprintf
2464         statements.
2465         * config/mcore/mcore-protos.h: Update prototypes for changed
2466         functions in mcore.c.
2467         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2468         HOST_WIDE_INT and not int.
2469         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2470         CONST_OK_FOR_N): Likewise.
2471         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2472         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2473         an INTVAL.
2474
2475 2007-01-10  Jan Hubicka  <jh@suse.cz>
2476
2477         * tree-vrp.c (remove_range_assertions): Release defs.
2478         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2479         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2480
2481 2007-01-10  Paul Brook  <paul@codesourcery.com>
2482
2483         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2484         * config/arm/arm.md (mulsi3subsi): New insn.
2485
2486 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
2487
2488         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2489         (tree_transform_and_unroll_loop): New.
2490         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2491         Declare.
2492
2493 2007-01-10  Robert Kennedy <jimbob@google.com>
2494
2495         * fold-const.c (fold_comparison): Fold comparisons like (x *
2496         1000 < 0) to (x < 0).
2497
2498 2007-01-10  Ian Lance Taylor  <iant@google.com>
2499
2500         * tree-pretty-print.c (dump_generic_node): Print parentheses when
2501         operands have the same priority.
2502
2503 2007-01-10  Tom Tromey  <tromey@redhat.com>
2504
2505         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2506         (fold_binary): Likewise.
2507         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2508         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2509         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2510         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2511         * langhooks.h (struct lang_hooks): Removed field
2512         'can_use_bit_fields_p'.
2513
2514 2007-01-10  Ralf Cors├ępius <ralf.corsepius@rtems.org>
2515
2516         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2517
2518 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
2519
2520         * function.c (get_last_funcdef_no): New function.
2521         * function.h (get_last_funcdef_no): Declare.
2522         * tree-inline.c (initialize_cfun): Add initialization.
2523         (tree_function_versioning): Cleanup.
2524
2525 2007-01-10  Jan Hubicka  <jh@suse.cz>
2526
2527         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2528         PHIs.
2529
2530 2007-01-10  Sa Liu  <saliu@de.ibm.com>
2531             Ben Elliston  <bje@au.ibm.com>
2532
2533         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2534         for SAVE_NONLOCAL and Pmode for any other save level.
2535         * spu-protos.h (spu_restore_stack_block): Declare.
2536         * spu.md (save_stack_block): Remove.
2537         (restore_stack_block): Call spu_restore_stack_block.
2538         * spu.c (spu_restore_stack_block): New function.
2539         (spu_expand_epilogue): Remove old comment.
2540
2541 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
2542
2543         PR tree-optimization/30322
2544         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2545         (cand_value_at): Return the value as aff_tree.
2546         (may_eliminate_iv): Convert the bound from aff_tree to tree.
2547         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2548         aff_combination_mult): New functions.
2549         (aff_combination_add): Use aff_combination_add_cst.
2550         (aff_combination_convert): Allow conversions to a wider type.
2551         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2552         * tree-affine.h (aff_combination_mult): Declare.
2553
2554 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
2555
2556         * doc/tm.texi: Update documentation to reflect reality of exec
2557         and start file search behaviours. Update copyright year.
2558         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2559         header file directories.
2560
2561 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
2562
2563         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2564         (*cosxf2): Rename to cosxf2_i387.
2565         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
2566         using X87MODEF12 mode macro. Rename patterns to
2567         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
2568         for SSE math.
2569         (*cosdf2, *cossf2): Ditto.
2570         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
2571         to match extended input operands.
2572         (sincos<mode>3): New expander.
2573         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2574         insn patterns and corresponding splitters.
2575
2576 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
2577
2578         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2579         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2580
2581 2007-01-09  Nicolas Pitre  <nico@cam.org>
2582
2583         PR target/30173
2584         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2585
2586 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2587
2588         * target.h (struct gcc_target): New field library_rtti_comdat.
2589         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2590         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2591         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2592         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2593
2594 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2595
2596         * doc/invoke.texi (Optimize Options): Correct description of -O0.
2597
2598 2007-01-08  Richard Guenther  <rguenther@suse.de>
2599
2600         * tree.h (force_fit_type_double): Export.
2601         (force_fit_type): Remove.
2602         * fold-const.c (force_fit_type_double): New function.
2603         (force_fit_type): Remove.
2604         (int_const_binop): Use it.
2605         (fold_convert_const_int_from_int): Likewise.
2606         (fold_convert_const_int_from_real): Likewise.
2607         (fold_div_compare): Likewise.
2608         (fold_sign_changed_comparison): Likewise.
2609         (fold_unary): Likewise.
2610         (fold_negate_const): Likewise.
2611         (fold_abs_const): Likewise.
2612         (fold_not_const): Likewise.
2613         * c-common.c (shorten_compare): Use force_fit_type_double.
2614         * convert.c (convert_to_pointer): Likewise.
2615
2616 2007-01-08  Richard Guenther  <rguenther@suse.de>
2617
2618         * tree.h (build_int_cst_wide_type): Export.
2619         * tree.c (build_int_cst_wide_type): New function.
2620         (build_int_cst_wide): Fix comment.
2621         * builtins.c (fold_builtin_object_size): Use build_int_cst
2622         to build -1 or 0 of the correct type.  Use fit_double_type
2623         to check for overflow.
2624         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2625         to build the mask.
2626         (decode_field_reference): Likewise.
2627         (all_ones_mask_p): Likewise.
2628         (native_interpret_int): Use build_int_cst_wide_type.
2629         (fold_binary): Use build_int_cst_type to build an all-ones
2630         value.
2631         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2632
2633 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
2634
2635         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2636
2637 2007-01-09  Ben Elliston  <bje@au.ibm.com>
2638
2639         * genautomata.c (STATS_OPTION): New option.
2640         (stats_flag): New flag.
2641         (gen_automata_option): Handle it.
2642         (initiate_automaton_gen): Ditto.
2643         (write_automata): Output statistics only if stats_flag is
2644         set. Likewise, output time statistics only if time_flag is set.
2645         * doc/md.texi (Processor pipeline description): Document new flag.
2646
2647 2007-01-08  Richard Guenther  <rguenther@suse.de>
2648
2649         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2650         * tree.c (build_int_cst_type): Likewise.
2651         (size_in_bytes): Don't call force_fit_type on the result.
2652         (int_fits_type_p): Use fit_double_type.
2653         * fold-const.c (fit_double_type): New function.
2654         (force_fit_type): Use it.
2655         * tree.h (fit_double_type): Export.
2656
2657 2007-01-08  Jan Hubicka  <jh@suse.cz>
2658
2659         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2660         * ipa.c (function_and_variable_visibility): Fix return type.
2661
2662 2007-01-08  Richard Guenther  <rguenther@suse.de>
2663
2664         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2665         of offset to build the index.
2666         * tree-pretty-print.c (dump_generic_node): Don't build negated
2667         const just for printing.
2668         * c-pretty-print.c (pp_c_integer_constant): Likewise.
2669         * builtins.c (fold_builtin_int_roundingfn): Check if result
2670         fits the type by using force_fit_type and comparing the result.
2671         * predict.c (predict_loops): Use compare_tree_int for comparison.
2672         * tree.c (build_int_cst): Fall back to integer_type_node for
2673         NULL_TREE type.
2674         (build_int_cst_wide): Assert type is non-null.
2675
2676 2007-01-08  Roberto Costa  <roberto.costa@st.com>
2677
2678         * tree-vrp.c (extract_range_from_cond_expr): New.
2679         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2680         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2681         as expressions.
2682         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2683         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2684         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2685         used as expressions.
2686         * tree-object-size.c (cond_expr_object_size): New.
2687         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2688
2689 2007-01-08  Jan Hubicka  <jh@suse.cz>
2690
2691         * tree-ssa-forwprop.c (forward_propagate_into_cond,
2692         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2693         statement.
2694
2695 2007-01-08  Richard Guenther  <rguenther@suse.de>
2696
2697         PR tree-optimization/23603
2698         * tree-vrp.c (set_value_range_to_truthvalue): New function.
2699         (extract_range_from_binary): Fall back to truthvalue instead of
2700         varying for TRUTH_*_EXPR.
2701         (extract_range_from_comparison): Fall back to truthvalue instead of
2702         varying.
2703         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2704         if all visited PHI values were constant.
2705
2706 2007-01-08  Jan Hubicka  <jh@suse.cz>
2707
2708         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2709         (cgraph_analyze_function): break out from ...
2710         (cgraph_finalize_compilation_unit): ... here.
2711         (cgraph_expand_function): Remove forgoten commented out line.
2712         (cgraph_optimize): Analyze functions.
2713
2714 2007-01-08  Jan Hubicka  <jh@suse.cz>
2715
2716         * tree-pas.h (TODO_remove_function): New flag.
2717         (TODO_update*): Renumber.
2718         (pass_ipa_increase_alignment,
2719         pass_ipa_function_and_variable_visibility): New passes.
2720         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2721         (cgraph_function_and_variable_visibility): Move to ipa.c
2722         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2723         cgraph_increase_alignment.
2724         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2725         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2726         calling tree-inline.
2727         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2728         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2729         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2730         (gate_increase_alignment): New function.
2731         (pass_ipa_increase_alignment): New pass.
2732         * ipa.c: Inline tree-pass.h and timevar.h
2733         (function_and_variable_visibility): Move here from cgraphunit.c
2734         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2735         * passes.c (init_optimization_passes): Add the two new passes.
2736         (execute_todo): Handle cgraph_remove_functions.
2737
2738 2007-01-08  Nick Clifton  <nickc@redhat.com>
2739
2740         * config/frv/predicates.md (reg_or_0_operand): Accept
2741         CONST_DOUBLEs.
2742
2743 2007-01-08  Ralf Cors├ępius <ralf.corsepius@rtems.org>
2744
2745         * config/bfin/rtems.h, config/bfin/t-rtems: New.
2746         * config.gcc: Add bfin*-rtems*.
2747
2748 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2749
2750         * c.opt: Add -flax-vector-conversions.
2751         * c-typeck.c (convert_for_assignment): Pass flag to
2752         vector_types_convertible_p to allow emission of note.
2753         (digest_init): Likewise.
2754         * c-opts.c: Handle -flax-vector-conversions.
2755         * c-common.c (flag_lax_vector_conversions): New.
2756         (vector_types_convertible_p): Unless -flax-vector conversions
2757         has been passed, disallow conversions between vectors with
2758         differing numbers of subparts and/or element types.  If such
2759         a conversion is disallowed, possibly emit a note on the first
2760         occasion only to inform the user of -flax-vector-conversions.
2761         The new last argument specifies this.
2762         * c-common.h (flag_lax_vector_conversions): New.
2763         (vector_types_convertible_p): Add extra argument.
2764         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2765         char_type_node for V*QI type vectors.
2766         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2767         Update to satisfy new typechecking rules.
2768         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2769         C and C++ variants.
2770         * doc/invoke.texi (C Dialect Options): Document
2771         -flax-vector-conversions.
2772
2773 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2774
2775         PR tree-optimization/29877
2776         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2777         a register variable on the RHS to not be replaceable.
2778
2779 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
2780         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2781         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2782         const_simm15, const_pow2, const_npow2): Added.
2783         * config/score/misc.md (insv, extv, extzv, movmemsi,
2784         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2785         * config/score/score.c (score_address_cost, score_select_cc_mode):
2786         Added.
2787         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2788         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2789         Update score_print_operand.
2790         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2791         Adjust register allocate order and update some macro define.
2792         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2793         mdx_block_move_straight, mdx_block_move_loop_head,
2794         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2795         mdx_block_move): Added.
2796         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2797         mdp_limm): Updated and fix some bug and typo.
2798         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2799         (movsf, movdf, doloop_end): Added.
2800
2801 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
2802
2803         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2804         config/arm/thumb2.md: Fix comment typos.
2805         * doc/extend.texi: Fix a typo.
2806
2807 2007-01-07  Eric Christopher  <echristo@apple.com>
2808
2809         * configure.ac: Check for __stack_chk_fail for darwin.
2810         * configure: Regenerate.
2811
2812 2007-01-07  Richard Guenther  <rguenther@suse.de>
2813
2814         * tree-vrp.c (extract_range_from_assert): CSE calls to
2815         compare_values where possible.
2816         (extract_range_from_unary_expr): Likewise.
2817
2818 2007-01-07  Anatoly Sokolov <aesok@post.ru>
2819
2820         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2821         * config/avr/avr.c (call_insn_operand): Delete function.
2822         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2823         instead of register number. Use predicates.
2824         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2825         call_insn_operand): Add.
2826
2827 2007-01-06  Jan Hubicka  <jh@suse.cz>
2828
2829         * tree-pass.h (pass_build_cgraph_edges): Declare.
2830         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2831         (visited_nodes): Remove.
2832         (cgraph_create_edges): Move to cgraphbuild.c; rename to
2833         build_cgrpah_edges; make visited_nodes local.
2834         (cgraph_process_new_functions): DO not call initialize_inline_failed.
2835         (record_references_in_initializer): Move to cgraphbuild.c
2836         (initialize_inline_failed, rebuild_cgraph_edges,
2837         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2838         (verify_cgraph_node): Make visited_nodes local.
2839         (cgraph_analyze_function): Do not call cgraph_create_edges and
2840         initialize_inline_failed.
2841         (cgraph_expand_function): Do not call cgraph_lower_function;
2842         assert that function is already lowered.
2843         * Makefile.in (cgraphbuild.o): New.
2844         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2845         at the end of lowering passes.
2846
2847 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
2848
2849         * ifcvt.c (cond_move_convert_if_block): New function, code
2850         factored out from...
2851         (cond_move_process_if_block): ...here.  Call the new function
2852         on the THEN and ELSE blocks.
2853         (merge_if_block): Do not copy global_live_at_end, merge_blocks
2854         already takes care of this.
2855
2856 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2857
2858         PR c/19978
2859         * tree.h (TREE_OVERFLOW_P): New.
2860         * c-typeck.c (parser_build_unary_op): Warn only if result
2861         overflowed and operands did not.
2862         (parser_build_binary_op): Likewise.
2863         (convert_for_assignment): Remove redundant overflow_warning.
2864         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2865
2866 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2867
2868         * c-typeck.c (store_init_value): Split over two lines to follow
2869         the GNU coding style.
2870
2871 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
2872
2873         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2874         __GXX_EXPERIMENTAL_CXX0X__.
2875         * doc/cpp.texi: Same.
2876
2877 2007-01-05  Richard Guenther  <rguenther@suse.de>
2878
2879         PR middle-end/27826
2880         * tree.c (get_narrower): Do not construct COMPONENT_REFs
2881         with mismatched types.  Instead explicitly build a
2882         conversion NOP_EXPR.
2883
2884 2007-01-05  Ian Lance Taylor  <iant@google.com>
2885
2886         * c-common.c (decl_with_nonnull_addr_p): New function.
2887         (c_common_truthvalue_conversion): Call it.
2888         * c-typeck.c (build_binary_op): Likewise.
2889         * c-common.h (decl_with_nonnull_addr_p): Declare.
2890
2891 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
2892
2893         PR c/30360
2894         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2895         denom against 0.0.
2896
2897 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
2898
2899         * doc/install.texi (Final install): Document the fact that
2900         the GNAT runtime should not be stripped.
2901
2902 2007-01-04  Jan Hubicka  <jh@suse.cz>
2903
2904         * tree-inline.c (fold_marked_statements): Update operand caches
2905         and EH after folding
2906
2907 2007-01-04  Ian Lance Taylor  <iant@google.com>
2908
2909         * c-common.c (check_function_nonnull): Whitespace fix.
2910
2911 2007-01-04  Jan Hubicka  <jh@suse.cz>
2912
2913         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2914         comitted older version of patch.
2915         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2916         function, celanup cfg and collect garbage.
2917
2918 2007-01-04  Mike Stump  <mrs@apple.com>
2919
2920         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2921         (clean): Likewise.
2922
2923 2007-01-04  Eric Christopher  <echristo@apple.com>
2924
2925         * libgcc2.c (__bswapsi2): Use SItype.
2926         (__bswapdi2): Use DItype.
2927         * libgcc2.h: Update for above.
2928
2929 2007-01-04  Paul Brook  <paul@codesourcery.com>
2930
2931         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2932         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2933         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2934         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2935         umulsi3_highpart): Make conditional on !arm_arch6.
2936         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2937         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2938         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2939         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2940         umulsi3_highpart_v6): New insns.
2941
2942 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2943
2944         * fold-const.c (fold_convert): When casting an expression to void,
2945         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2946         have a type.  Instead of attempting to build a NOP_EXPR, return
2947         these "special" trees directly.
2948
2949 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2950
2951         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2952         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2953         rs6000_cost->fp not 0 as adjustment for outer NEG.
2954
2955 2007-01-04  Jan Hubicka  <jh@suse.cz>
2956
2957         * cgraph.c (cgraph_release_function_body): New function.
2958         (cgraph_remove_node): Use it.
2959         * cgraph.h (cgraph_release_function_body): Declare.
2960         * cgraphunit.c (cgraph_expand_function): Use it.
2961         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2962         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2963         is initialized and while compilation of other function is running.
2964         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2965         statement CFG annotations from here to ...
2966         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2967
2968 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2969
2970         * cfgloop.h (enum li_flags): Make the constants powers of two.
2971
2972 2007-01-04  Jan Hubicka  <jh@suse.cz>
2973
2974         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2975         set.
2976         (fold_marked_statements): New function.
2977         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2978         * tree-inline.h (copy_body_data): Add statements_to_fold.
2979
2980 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2981
2982         * config.gcc: Mention libgcc/config.host.
2983         * Makefile.in: Update comments mentioning libgcc.
2984         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2985         (all.cross, start.encap, rest.encap, rest.cross): Update
2986         dependencies for libgcc move.
2987         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2988         (clean-target, clean-target-libgcc): Delete.
2989         (srcdirify, GCC_EXTRA_PARTS): New macros.
2990         (libgcc-support, libgcc.mvars): New rules.
2991         (distclean): Remove mention of mklibgcc.
2992         (install): Don't reference INSTALL_LIBGCC.
2993         (install-common): Don't reference EXTRA_PARTS.
2994         (install-libgcc, install-multilib): Delete rules.
2995         * mklibgcc.in: Delete file.
2996         * doc/configfiles.texi: Don't mention mklibgcc.
2997
2998         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2999         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
3000         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
3001         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
3002         (mklibgcc, ldblspecs): Likewise.
3003
3004         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
3005         (posixpre.def): Use $(T).
3006         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
3007         (SHLIB_SRC, SHLIB_INSTALL): Delete.
3008         (SHLIB_LINK): Make dummy.
3009         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
3010
3011         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
3012
3013         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
3014         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
3015         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
3016         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
3017
3018         * configure.ac (all_outputs): Remove mklibgcc.
3019         * configure: Regenerated.
3020
3021 2007-01-03  Josh Conner  <jconner@apple.com>
3022
3023         PR middle-end/29683
3024         * calls.c (compute_argument_addresses): Set stack and stack_slot
3025         for partial args, too.
3026         (store_one_arg): Use locate.size.constant for the size when
3027         generating a save_area.
3028
3029 2007-01-03  Robert Kennedy <jimbob@google.com>
3030
3031         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
3032         whose definitions are deleted due to basic block merging.
3033
3034 2007-01-03  Paul Brook  <paul@codesourcery.com>
3035
3036         PR target/16634
3037         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
3038         functions.
3039         (use_return_insn): Return 0 for Thumb interrupt functions.
3040         (print_multi_reg): Add rfe argument for IRQ returns.
3041         (arm_output_epilogue): Pop interrupt return address directly into PC.
3042         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
3043
3044 2007-01-03  Paul Brook  <paul@codesourcery.com>
3045
3046         Merge from sourcerygxx-4_1.
3047         * config/arm/thumb2.md: New file.
3048         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
3049         Thumb-2.
3050         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
3051         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3052         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
3053         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
3054         tables.
3055         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3056         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
3057         comptibility.
3058         * config/arm/ieee754-sf.S: Ditto.
3059         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
3060         (thumb1_base_register_rtx_p): ... to this.
3061         (thumb_index_register_rtx_p): Rename...
3062         (thumb1_index_register_rtx_p): ... to this.
3063         (thumb_output_function_prologue): Rename...
3064         (thumb1_output_function_prologue): ... to this.
3065         (thumb_legitimate_address_p): Rename...
3066         (thumb1_legitimate_address_p): ... to this.
3067         (thumb_rtx_costs): Rename...
3068         (thumb1_rtx_costs): ... to this.
3069         (thumb_compute_save_reg_mask): Rename...
3070         (thumb1_compute_save_reg_mask): ... to this.
3071         (thumb_final_prescan_insn): Rename...
3072         (thumb1_final_prescan_insn): ... to this.
3073         (thumb_expand_epilogue): Rename...
3074         (thumb1_expand_epilogue): ... to this.
3075         (arm_unwind_emit_stm): Rename...
3076         (arm_unwind_emit_sequence): ... to this.
3077         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3078         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3079         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3080         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3081         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3082         thumb2_output_casesi): New functions.
3083         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3084         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3085         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3086         THUMB2_WORK_REGS): Define.
3087         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3088         arm_condexec_mask, arm_condexec_masklen)): New variables.
3089         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3090         (arm_override_options): Check new CPU capabilities.
3091         Set new architecture flag variables.
3092         (arm_isr_value): Handle v7m interrupt functions.
3093         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
3094         Thumb-2.
3095         (const_ok_for_arm): Handle Thumb-2 constants.
3096         (arm_gen_constant): Ditto.  Use movw when available.
3097         (arm_function_ok_for_sibcall): Return false for v7m interrupt
3098         functions.
3099         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3100         (thumb_find_work_register, arm_load_pic_register,
3101         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3102         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3103         print_multi_reg, output_mov_long_double_fpa_from_arm,
3104         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3105         output_mov_double_fpa_from_arm, output_move_double,
3106         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3107         output_return_instruction, arm_output_function_prologue,
3108         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3109         arm_output_mi_thunk, thumb_set_return_address): Ditto.
3110         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
3111         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3112         (arithmetic_instr, shift_op): Use arm_shift_nmem.
3113         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
3114         '!' and 'L'.
3115         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3116         (thumb_expand_prologue): Use thumb_set_frame_pointer.
3117         (arm_file_start): Output directive for unified syntax.
3118         (arm_unwind_emit_set): Handle stack alignment instruction.
3119         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3120         Add v6t2, v7, v7a, v7r and v7m.
3121         (RETLDM): Add Thumb-2 code.
3122         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3123         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3124         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3125         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3126         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3127         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3128         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3129         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3130         ADJUST_INSN_LENGTH): Define.
3131         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3132         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3133         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3134         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3135         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3136         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3137         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3138         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3139         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3140         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3141         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3142         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3143         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3144         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3145         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3146         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3147         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3148         cortex-m3.
3149         * config/arm/arm-tune.md: Regenerate.
3150         * config/arm/arm-protos.h: Update prototypes.
3151         * config/arm/vfp.md: Enable patterns for Thumb-2.
3152         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
3153         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3154         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3155         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3156         * config/arm/libunwind.S: Add Thumb-2 code.
3157         * config/arm/constraints.md: Update include Thumb-2.
3158         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3159         * config/arm/ieee754-df.S: Ditto.
3160         * config/arm/bpabi.S: Ditto.
3161         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3162         * config/arm/predicates.md (low_register_operand,
3163         low_reg_or_int_operand, thumb_16bit_operator): New.
3164         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3165         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3166         * config/arm/t-arm-elf: Add armv7 multilib.
3167         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3168         Include thumb2.md.
3169         (UNSPEC_STACK_ALIGN, ce_count): New.
3170         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3171         arm_zero_extendsidi2, arm_zero_extendqidi2): New
3172         insns/expanders.
3173         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3174         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3175         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3176         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3177         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3178         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3179         insns.
3180         * doc/extend.texi: Document ARMv7-M interrupt functions.
3181         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3182
3183 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
3184
3185         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3186         (struct _Unwind_Context): Rename args_size to flags, remove
3187         signal_frame field, add a new args_size field and version field.
3188         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3189         _Unwind_IsExtendedContext): New inline functions.
3190         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3191         Assume by_value array is only present if _Unwind_IsExtendedContext.
3192         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3193         _Unwind_IsSignalFrame.
3194         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3195         (uw_update_context_1): Use _Unwind_SetSignalFrame.
3196         (uw_init_context_1): Initialize context->flags to
3197         EXTENDED_CONTEXT_BIT.
3198         * config/rs6000/linux-unwind.h (frob_update_context): Use
3199         _Unwind_SetSignalFrame.
3200
3201 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3202
3203         PR middle-end/30353
3204         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3205         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3206         the value.
3207
3208 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3209
3210         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3211         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3212         RET, AT_SP): Remove.
3213         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3214         instead of RET.
3215
3216         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3217         * cfgbuild.c (find_basic_blocks): Likewise.
3218         * cfgrtl.c (rtl_create_basic_block): Likewise.
3219         * function.c (temp_slots_at_level): Likewise.
3220         * reg-stack.c (stack_regs_mentioned): Likewise.
3221         * regclass.c (allocate_reg_info): Likewise.
3222         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3223         set_bb_for_stmt, move_block_to_fn): Likewise.
3224         * tree-complex.c (tree_lower_complex): Likewise.
3225         * vec.h (VEC_safe_grow_cleared): New.
3226
3227         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3228
3229 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
3230
3231         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3232         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3233         to loop_version.
3234         * cfgloopmanip.c (scale_loop_frequencies): Export.
3235         (loopify): Scale the frequencies by prescribed coefficients.
3236         (set_zero_probability): New function.
3237         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3238         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3239         and frequencies according to arguments.
3240         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3241         correctly.
3242         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3243         * modulo-sched.c (sms_schedule): Set probabilities for entering
3244         versioned loop correctly.
3245         * tree-vect-transform.c (vect_transform_loop): Ditto.
3246         * cfgloop.h (loopify, loop_version): Declaration changed.
3247         (scale_loop_frequencies): Declared.
3248
3249 2007-01-02  Jan Hubicka  <jh@suse.cz>
3250
3251         * cgraph.c: Include tree-flow.h
3252         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3253         early_local_passes.
3254         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3255         * tree-pass.h (pass_all_early_optimizations): Declare.
3256         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3257         early_local_passes.
3258         (cgraph_analyze_function): Do early_local_passes.
3259         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3260         Do not add referenced vars.
3261         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3262         SSA form.
3263         (gate_all_early_local_passes): New gate.
3264         (pass_early_local_passes): Use new gate.
3265         (execute_early_local_optimizations): New functions.
3266         (gate_all_early_optimizations): New gate.
3267         (pass_all_early_optimizations): New pass.
3268         (execute_free_datastructures): Free SSA only when initialized.
3269         (gate_init_datastructures): Init only when optimizing.
3270         (tree_lowering_passes): Do early local passes when called late.
3271         * tree-profile.c (do_tree_profiling): Don't profile functions added
3272         late.
3273         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3274         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3275         not active.
3276         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3277         during early_local_passes.
3278         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3279
3280
3281 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3282
3283         * Makefile.in: Update copyright year.
3284
3285 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3286
3287         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3288         in $(lang_checks) and check-consistency targets.
3289
3290 2007-01-02  Jan Hubicka  <jh@suse.cz>
3291
3292         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3293         Do not add referenced vars.
3294         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3295         are not active.
3296         * passes.c (init_optimization_passes): Put mudflap_2 after
3297         free_datastructures.
3298
3299 2007-01-02  Jan Hubicka  <jh@suse.cz>
3300
3301         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3302         Set NOTHROW flag on call statements proved to be nothrow.
3303         Update statement of local calls so new pure/const functions are
3304         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3305         goto receivers.
3306         (tree_rest_of_compilation): Register hooks and initialize bitmap
3307         early. Do not set after_inlining flag.
3308
3309 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3310
3311         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3312         HOST_BITS_PER_LONGLONG
3313
3314 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3315
3316         PR c/19977
3317         * c-typeck.c (store_init_value): Don't emit pedantic overflow
3318         warning for non-static initializers.
3319
3320 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
3321
3322         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3323         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3324         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3325         Always use set_unique_reg_note to add REG_EQUAL notes.
3326
3327 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3328
3329         Revert:
3330         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3331
3332         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3333         * cfgbuild.c (find_basic_blocks): Likewise.
3334         * cfgrtl.c (rtl_create_basic_block): Likewise.
3335         * function.c (temp_slots_at_level): Likewise.
3336         * reg-stack.c (stack_regs_mentioned): Likewise.
3337         * regclass.c (allocate_reg_info): Likewise.
3338         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3339         set_bb_for_stmt, move_block_to_fn): Likewise.
3340         * tree-complex.c (tree_lower_complex): Likewise.
3341         * vec.h (VEC_safe_grow_cleared): New.
3342
3343 2007-01-02  Ian Lance Taylor  <iant@google.com>
3344
3345         * c-common.c (c_common_truthvalue_conversion): When warning about
3346         using an assignment as a truth value, set TREE_NO_WARNING.
3347
3348 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3349
3350         PR middle-end/7651
3351         * c.opt (Wold-style-declaration): New.
3352         * doc/invoke.texi (C-only Warning Options): New.
3353         (Wold-style-declaration): Document it.
3354         (Wextra): Enabled by -Wextra.
3355         * c-opts.c (c_common_post_options): Enabled by -Wextra.
3356         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3357         -Wold-style-declaration.
3358
3359 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3360
3361         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3362         * cfgbuild.c (find_basic_blocks): Likewise.
3363         * cfgrtl.c (rtl_create_basic_block): Likewise.
3364         * function.c (temp_slots_at_level): Likewise.
3365         * reg-stack.c (stack_regs_mentioned): Likewise.
3366         * regclass.c (allocate_reg_info): Likewise.
3367         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3368         set_bb_for_stmt, move_block_to_fn): Likewise.
3369         * tree-complex.c (tree_lower_complex): Likewise.
3370         * vec.h (VEC_safe_grow_cleared): New.
3371
3372 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
3373
3374         * c-common.c (c_common_nodes_and_builtins): Since variants of
3375         void_type_node get built before it is given a name, we need to
3376         give those variants the name, too.
3377         (complete_array_type): We need to work with the canonical main
3378         type of the array, from which we will build the qualified version.