1 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
3 * pointer-set.c (insert_aux): Only return insertion slot.
4 (pointer_set_insert): Adjust.
5 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
6 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
7 pointer_map_traverse): New.
8 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
9 pointer_map_create, pointer_map_destroy, pointer_map_insert,
10 pointer_map_contains, pointer_map_traverse): Declare.
12 2007-02-02 Jakub Jelinek <jakub@redhat.com>
15 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
16 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
18 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
20 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
22 (struct _dep): New type.
24 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
25 (DEP_STATUS): New access macro. The macro with the same name was
26 renamed to DEP_LINK_STATUS.
27 (dep_init): Declare function
29 (struct _dep_link): New type.
30 (dep_link_t): New typedef.
31 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
32 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
33 (DEP_LINK_STATUS): New macro.
34 (debug_dep_links): New debug function.
36 (struct _deps_list): New type.
37 (deps_list_t): New typedef.
38 (DEPS_LIST_FIRST): New access macro.
39 (FOR_EACH_DEP_LINK): New cycle macro.
40 (create_deps_list, free_deps_list, delete_deps_list): Declare
42 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
43 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
44 (copy_deps_list_change_con): Ditto.
46 (move_dep_link): Declare function.
48 (struct _dep_node): New type.
49 (dep_node_t): New typedef.
50 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
52 (struct haifa_insn_data.back_deps): New field to hold backward
53 dependencies of the insn.
54 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
56 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
57 Change its type to deps_list_t.
58 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
59 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
60 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
62 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
64 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
66 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
70 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
71 (ds_to_dk, dk_to_ds): New functions.
73 (init_dep_1): New static function.
74 (init_dep): New function.
75 (copy_dep): New static function.
77 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
79 (detach_dep_link): New static function.
80 (move_dep_link): New function.
82 (dep_links_consistent_p, dump_dep_links): New static functions.
83 (debug_dep_links): New debugging function.
85 (deps_obstack, dl_obstack, dn_obstack): New static variables.
87 (alloc_deps_list, init_deps_list): New static functions.
88 (create_deps_list): New function.
89 (clear_deps_list): New static function.
90 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
91 (deps_list_consistent_p, dump_deps_list): New static functions.
92 (debug_deps_list): New function.
93 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
95 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
97 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
98 use new scheduler dependencies lists.
99 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
100 (sched_analyze): Ditto. Initialize dependencies lists.
101 (add_forw_dep, compute_forward_dependences): Update to use new
102 scheduler dependencies lists.
104 (init_dependency_caches): Init deps_obstack.
105 (free_dependency_caches): Free deps_obstack.
107 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
108 new scheduler dependencies lists.
109 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
110 (add_back_forw_dep, delete_back_forw_dep): Ditto.
112 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
113 Update to use new scheduler dependencies lists.
114 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
115 (debug_dependencies): Ditto.
116 (schedule_region): Update comments.
118 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
119 scheduler dependencies lists.
120 (schedule_ebb): Update comments.
122 * rtl.def (DEPS_LIST): Remove.
124 * lists.c (unused_deps_list): Remove.
125 (free_list): Update assertions.
127 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
128 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
130 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
131 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
133 * haifa-sched.c (comments): Update.
134 (insn_cost1): Remove. Inline the code into insn_cost ().
135 (insn_cost): Update to use new scheduler dependencies lists. Move
136 processing of the dependency cost to dep_cost ().
137 (dep_cost): New function. Use it instead of insn_cost () when
138 evaluating cost of the dependency. Use compatible interface to
139 interact with the target.
140 (priority): Update to use new scheduler dependencies lists.
141 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
142 with greater number of insns that depend on the insn.
143 (schedule_insn): Update to use new scheduler dependencies lists. Add
144 code to free backward dependencies lists. Inline and optimize code
145 from resolve_dep () - see PR28071.
146 (ok_for_early_queue_removal): Update to use new scheduler dependencies
147 lists. Update call to targetm.sched.is_costly_dependence hook.
149 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
150 scheduler dependencies lists.
152 (resolve_dep): Remove. Move the logic to schedule_insn ().
153 (init_h_i_d): Initialize dependencies lists.
155 (process_insn_depend_be_in_spec): Rename to
156 process_insn_forw_deps_be_in_spec. Update to use new scheduler
158 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
159 Update to use new scheduler dependencies lists.
160 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
162 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
163 new scheduler dependencies lists.
164 (build_intra_loop_deps): Ditto.
166 * target.h (struct _dep): Declare to use in
167 gcc_target.sched.is_costly_dependence.
168 (struct gcc_target.sched.adjust_cost): Fix typo.
169 (struct gcc_target.sched.is_costly_dependence): Change signature to use
170 single dep_t parameter instead of an equivalent triad.
171 (struct gcc_target.sched.adjust_cost_2): Remove.
173 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
175 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
177 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
179 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
181 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
182 (REG_DEP_TRUE): Document.
184 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
185 Change signature to correspond to the targetm.sched.adjust_cost hook.
186 Update use in TARGET_SCHED_ADJUST_COST_2.
187 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
188 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
189 new scheduler dependencies lists.
190 (ia64_gen_check): Ditto.
192 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
195 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
196 signature to correspond to the targetm.sched.is_costly_dependence hook.
197 (is_costly_group): Update to use new scheduler dependencies lists.
199 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
200 instead of INSN_COST () macro.
202 2007-02-01 Ian Lance Taylor <iant@google.com>
204 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
206 2007-02-01 Guy Martin <gmsoft@gentoo.org>
208 * pa.md (tp_load): Correct mfctl instruction syntax.
210 2007-02-01 Geoffrey Keating <geoffk@apple.com>
212 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
213 altivec_padding_size calculation on AIX. Improve comment, add
214 assert to verify that it's right.
216 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
219 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
222 * global.c (global_alloc): Add mising initialization of
225 2007-02-01 Roger Sayle <roger@eyesopen.com>
227 * alias.c (init_alias_analysis): Correct whitespace.
228 * bb-reorder.c (fix_edges_for_rarely_executed_code,
229 partition_hot_cold_basic_blocks): Likewise.
230 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
231 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
232 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
233 fold_builtin_snprintf_chk, fold_builtin_printf,
234 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
235 do_mpfr_arg3, do_mpfr_sincos): Likewise.
236 * cfgcleanup.c (cleanup_cfg): Likewise.
237 * cfgexpand.c (tree_expand_cfg): Likewise.
238 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
239 * function.c (get_next_funcdef_no): Likewise.
240 * gengtype.c (main): Likewise.
241 * genmodes.c (main): Likewise.
242 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
243 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
244 * ifcvt.c (noce_emit_move_insn): Likewise.
245 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
247 * stor-layout.c (get_best_mode): Likewise.
248 * tree-ssa-loop-niter.c (get_val_for): Likewise.
249 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
250 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
251 perform_var_substitution, solve_graph): Likewise.
252 * tree-vrp.c (vrp_finalize): Likewise.
254 2007-02-01 Ian Lance Taylor <iant@google.com>
256 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
257 high part of a paradoxical subreg, return a constant zero.
259 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
261 * toplev.c (lang_dependent_init): Call init_set_costs.
262 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
264 2007-02-01 Richard Guenther <rguenther@suse.de>
267 * fold-const.c (fold_negate_expr): Allow negating a
268 constant if overflow does not change.
270 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
272 * doc/c-tree.texi (Expression trees): Improve markup.
273 * doc/tm.texi (Register Classes, Addressing Modes)
274 (Floating Point): Fix spacing after abbreviations. Fix some
277 2007-02-01 Ben Elliston <bje@au.ibm.com>
279 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
280 * doc/contrib.texi: Likewise.
281 * doc/install.texi: Likewise.
283 2007-01-31 Richard Henderson <rth@redhat.com>
284 Ian Lance Taylor <iant@google.com>
286 * lower-subreg.c: New file.
287 * rtl.def (CONCATN): Define.
288 * passes.c (init_optimization_passes): Add pass_lower_subreg and
290 * emit-rtl.c (update_reg_offset): New static function, broken out
291 of gen_rtx_REG_offset.
292 (gen_rtx_REG_offset): Call update_reg_offset.
293 (gen_reg_rtx_offset): New function.
294 * regclass.c: Revert patch of 2006-03-05, restoring
296 (clear_reg_info_regno): New function.
297 * dwarf2out.c (concatn_loc_descriptor): New static function.
298 (loc_descriptor): Handle CONCATN.
299 * common.opt (fsplit_wide_types): New option.
300 * opts.c (decode_options): Set flag_split_wide_types when
302 * timevar.def (TV_LOWER_SUBREG): Define.
303 * rtl.h (gen_reg_rtx_offset): Declare.
304 (reg_scan_update): Declare.
305 * regs.h (clear_reg_info_regno): Declare.
306 * tree-pass.h (pass_lower_subreg): Declare.
307 (pass_lower_subreg2): Declare.
308 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
309 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
311 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
312 * Makefile.in (OBJS-common): Add lower-subreg.o.
313 (lower-subreg.o): New target.
315 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
317 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
319 2007-01-31 Anatoly Sokolov <aesok@post.ru>
322 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
324 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
327 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
329 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
331 * cfgloop.h: Include vec-prim.h.
332 (enum li_flags): Remove LI_ONLY_OLD.
333 (loop_iterator): Changed.
334 (fel_next, fel_init): Iterate over loop tree.
335 (FOR_EACH_LOOP_BREAK): New macro.
336 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
338 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
339 * modulo-sched.c (sms_schedule): Ditto.
340 * tree-vectorizer.c (vectorize_loops): Ditto.
341 * doc/loop.texi: Update information on loop numbering and behavior of
342 FOR_EACH_LOOP wrto new loops.
343 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
344 add_to_evolution_1): Test nestedness of loops instead of comparing
346 * tree-chrec.c (chrec_fold_plus_poly_poly,
347 chrec_fold_multiply_poly_poly, chrec_evaluate,
348 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
349 reset_evolution_in_loop): Ditto.
350 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
352 2007-01-31 Dirk Mueller <dmueller@suse.de>
354 * c-common.c (warn_about_parentheses): Separate warning about
355 un-parenthized sequence of comparison operators from the one
356 which is supposed to warn about x <= y <= z.
358 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
360 * optabs.h (enum optab_index): Add new OTI_isinf.
361 (isinf_optab): Define corresponding macro.
362 * optabs.c (init_optabs): Initialize isinf_optab.
363 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
365 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
366 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
368 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
369 expand_builtin_interclass_mathfn.
370 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
371 * config/i386/i386.md (UNSPEC_FXAM): New constant.
372 (fxam<mode>2_i387): New insn pattern.
373 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
374 built-in functions as x87 inline asm.
376 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
378 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
380 2007-01-30 Eric Christopher <echristo@apple.com>
382 * config.gcc: Add geode.
384 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
386 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
387 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
388 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
389 value-prof.c: Fix comment typos.
391 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
394 * doc/invoke.texi (Wpointer-arith): Document warning.
396 2007-01-30 Janis Johnson
398 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
399 the list of discrepancies from the draft TR.
401 2007-01-30 Dirk Mueller <dmueller@suse.de>
404 * doc/invoke.texi (-Wreturn-type): Update description to
407 2007-01-30 Richard Sandiford <richard@codesourcery.com>
409 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
411 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
413 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
414 to handle optabs that operate on floating point input argument and
415 output to integer output.
416 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
417 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
418 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
419 expand_builtin_interclass_mathfn ().
420 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
422 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
423 ilogbl built-in functions as x87 intrinsics.
425 2007-01-30 Richard Guenther <rguenther@suse.de>
428 * dwarf2out.c (reference_to_unused): Query varpool if the
431 2007-01-30 Richard Guenther <rguenther@suse.de>
434 * passes.c (execute_one_pass): Reset in_gimple_form to not
435 confuse non-unit-at-a-time mode.
437 2007-01-29 Roger Sayle <roger@eyesopen.com>
438 Richard Guenther <rguenther@suse.de>
440 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
441 avoid undefined behaviour on overflow. Use force_fit_type_double
442 to construct the constant with the specified TREE_OVERFLOW.
444 2007-01-29 Janis Johnson <janis187@us.ibm.com>
446 * config/dfp-bit.c: Add parameterized support for fp exceptions.
447 * config/dfp-bit.h: Ditto.
449 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
451 * c-decl.c (pop_scope): Replace warnings with call to
452 warn_for_unused_label.
453 * c-common.h (warn_for_unused_label): Declare.
454 * c-common.c (warn_for_unused_label): Define.
456 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
458 * tree-optimize.c (update_inlined_to_pointers): Delete unused
461 2007-01-29 Janis Johnson <janis187@us.ibm.com>
463 * Makefile.in (USER_H): Remove decfloat.h.
464 * ginclude/decfloat.h: Delete, moving contents to ...
465 * ginclude/float.h: Add support for decimal floating point,
466 guarded by __STDC_WANT_DEC_FP__.
468 2007-01-29 Mike Stump <mrs@apple.com>
470 * doc/gccint.texi (Top): Rename Loop Representation to Loop
471 Analysis and Representation to resolve case insensitive conflict.
472 * doc/loop.texi (Loop Analysis and Representation): Likewise.
474 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
476 * tree.h (struct tree_memory_tag): Add aliases member.
477 (MTAG_ALIASES): New macro.
478 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
479 (add_may_alias): Remove pointer-set. Update for may_aliases being
481 (mark_aliases_call_clobbered): Update for may_aliases being a
483 (compute_tag_properties): Ditto.
484 (create_partition_for): Ditto.
485 (compute_memory_partitions): Ditto.
486 (dump_may_aliases_for): Ditto.
487 (is_aliased_with): Ditto.
488 (add_may_alias_for_new_tag): Ditto.
489 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
490 (compute_is_aliased): New function.
491 (compute_may_aliases): Call compute_is_aliased).
492 (init_alias_info): Initialize alias_bitmap_obstack.
493 (union_alias_set_into): New function.
494 (compute_flow_sensitive_aliasing): Use union_aliases_into.
495 (have_common_aliases_p): Rewrite to take two bitmaps and use
497 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
499 (finalize_ref_all_pointers): Update for add_may_alias changes.
500 (new_type_alias): Ditto.
501 * tree-flow-inline.h (may_aliases): Return a bitmap.
502 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
503 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
504 may_aliases being a bitmap.
505 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
506 may_aliases now returns a bitmap.
507 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
509 * tree-ssa-operands.c (add_virtual_operand): Update for
510 may_aliases being a bitmap.
512 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
514 PR tree-optimization/30630
515 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
518 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
521 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
522 expanded only once even if an error occurs.
524 2007-01-29 Ben Elliston <bje@au.ibm.com>
526 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
527 (__gcov_average_profiler): Likewise.
528 (__gcov_ior_profiler): Likewise.
529 (__gcov_merge_ior): Likewise.
531 2007-01-28 Jan Hubicka <jh@suse.cz>
533 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
534 * value-prof.c (stringop_block_profile): Likewise.
536 2007-01-28 Jan Hubicka <jh@suse.cz>
538 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
539 variant handling histograms; add wrapper.
540 (clear_storage_via_libcall): Export.
541 (emit_block_move_hints): Break out from ...; add histograms.
542 (emit_block_move): ... this one.
543 (clear_storage_hints): Break out from ...; add histograms.
544 (clear_storage): ... this one.
545 (set_storage_via_memset): Handle histogram.
546 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
547 (clear_storage_hints, clear_storage_via_libcall): Declare.
548 (set_storage_via_setmem): Update prototype.
549 * doc/md.texi (movmem, setmem): Document new arguments.
551 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
553 (stringop_block_profile): New global function.
554 (tree_stringops_values_to_profile): Profile block size and alignment.
555 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
557 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
558 (stringop_block_profile): Declare.
559 * builtins.c: Include value-prof.h.
560 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
561 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
562 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
563 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
564 * profile.c (instrument_values): Add new counters.
565 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
567 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
568 (tree_init_edge_profiler): Build new profilers.
569 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
570 (pass_tree_profile): Add dump.
571 (tree_profile_hooks): Update.
572 * Makefile.in (LIBGCOV): Add new constants.
573 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
574 __gcov_ior_profiler): New.
575 * i386.md (movmem/setmem expanders): Add new optional arguments.
577 2007-01-28 David Edelsohn <edelsohn@gnu.org>
579 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
581 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
583 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
584 (tree_unswitch_single_loop): Pass eni_size_weights to
586 * tree-ssa-loop-manip.c: Include tree-inline.h.
587 (can_unroll_loop_p): Pass eni_size_weights to
589 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
590 Pass eni_size_weights to estimate_num_insns.
591 * tree.h (init_inline_once): Export.
592 * toplev.c (backend_init): Call init_inline_once.
593 * cgraphunit.c (cgraph_process_new_functions,
594 cgraph_analyze_function): Pass eni_inlining_weights to
596 * ipa-inline.c (compute_inline_parameters): Ditto.
597 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
598 to estimate_num_insns.
599 (try_unroll_loop_completely): Pass eni_size_weights to
601 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
602 ot estimate_num_insns.
603 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
604 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
605 * tree-inline.c (eni_inlining_weights, eni_size_weights,
606 eni_time_weights): New variables.
607 (init_inline_once): Initialize them.
608 (struct eni_data): Mew.
609 (estimate_num_insns_1, estimate_num_insns): Use weights.
610 * tree-inline.h (struct eni_weights_d): New.
611 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
612 (estimate_num_insns): Declaration changed.
613 * cfgloop.h (tree_num_loop_insns): Declaration changed.
614 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
615 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
617 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
619 * tree-data-ref.c (conflict_fn): Assert that the number of affine
620 relations in the conflict function is valid.
622 2007-01-27 Ian Lance Taylor <iant@google.com>
624 * common.opt: Add fstrict-overflow.
625 * opts.c (decode_options): Set flag_strict_overflow if -O2.
626 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
627 (TYPE_OVERFLOW_UNDEFINED): Define.
628 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
630 * tree.h (TYPE_TRAP_SIGNED): Don't define.
631 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
632 (fold_negate_expr): Likewise.
633 (make_range): Likewise.
634 (extract_muldiv_1): Likewise.
635 (maybe_canonicalize_comparison): Likewise.
636 (fold_comparison): Likewise.
637 (fold_binary): Likewise.
638 (tree_expr_nonnegative_p): Likewise.
639 (tree_expr_nonzero_p): Likewise.
640 * tree-vrp.c (compare_values): Likewise.
641 (extract_range_from_binary_expr): Likewise.
642 (extract_range_from_unary_expr): Likewise.
643 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
645 (nowrap_type_p): Likewise.
646 * tree-scalar-evolution.c (simple_iv): Likewise.
647 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
648 (build_range_check): Likewise.
649 (extract_muldiv_1): Likewise.
650 (fold_comparison): Likewise.
651 * tree-vrp.c (vrp_int_const_binop): Likewise.
652 (extract_range_from_unary_expr): Likewise.
653 * convert.c (convert_to_integer): Likewise.
654 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
655 (fold_comparison): Likewise.
656 (fold_binary): Likewise.
657 * optabs.c (optab_for_tree_code): Likewise.
658 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
659 * simplify-rtx.c (simplify_const_relational_operation): Check
660 flag_strict_overflow and flag_trapv.
661 (simplify_const_relational_operation): Likewise.
662 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
663 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
666 2007-01-27 Roger Sayle <roger@eyesopen.com>
668 * tree.c (tree_fold_gcd): Delete.
669 * tree.h (tree_fold_gcd): Remove prototype.
670 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
671 test whether one constant integer is a multiple of another. Instead
672 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
673 * fold-const.c (multiple_of_p): We've determined both TOP and
674 BOTTOM are integer constants so we can call int_const_binop directly
675 instead of the more generic const_binop.
677 2007-01-27 Roger Sayle <roger@eyesopen.com>
679 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
680 1*X check that the constant hasn't overflowed, to preserve the
682 (round_up): Provide an efficient implementation when rouding-up an
683 INTEGER_CST to a power-of-two.
685 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
687 * doc/sourcebuild.texi: Add comma for clarity.
688 * doc/extend.texi: Fix some typos.
689 * doc/passes.texi: Likewise.
690 * doc/cppinternals.texi: Likewise.
691 * doc/c-tree.texi: Likewise.
692 * doc/tree-ssa.texi: Likewise.
693 * doc/install.texi: Likewise.
695 2007-01-27 Jan Hubicka <jh@suse.cz>
697 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
698 (pass_sra): Do not require alias information.
699 * passes.c (init_optimization_passes): Add SRA
701 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
703 * tracer.c (rest_of_handle_tracer): We already cleaned
704 up the CFG in tracer() so don't do it here again.
705 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
706 cleanup_cfg here, either. And don't call renumber_insns.
708 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
709 and BB_HEAD are in the insn stream here. Instead make sure
710 that BB_INSN is valid on all insns. Also, do check here that
711 there are no pending branch predictions...
712 (rtl_verify_flow_info): ...instead of doing it here. Checks
713 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
716 2007-01-26 Roger Sayle <roger@eyesopen.com>
718 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
719 function to simplify/factorize operand order canonicalization.
720 (ix86_fixup_binary_operands): Reorganize using the above function.
721 (ix86_binary_operator_ok): Likewise.
723 2007-01-27 Jakub Jelinek <jakub@redhat.com>
725 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
726 file scope from optimize_attrs.
727 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
729 (optimize_attrs): Clear insn_code_values after freeing it.
731 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
733 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
734 a parts.base assignment.
736 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
738 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
739 (compute_subscript_distance, initialize_data_dependence_relation,
740 finalize_ddr_dependent, analyze_ziv_subscript,
741 analyze_siv_subscript_cst_affine,
742 compute_overlap_steps_for_affine_univar,
743 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
744 analyze_siv_subscript, analyze_miv_subscript,
745 analyze_overlapping_iterations, subscript_dependence_tester_1,
746 compute_self_dependence, free_dependence_relation): Work
747 with affine_fn instead of chrecs.
748 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
749 common_affine_function, affine_function_base,
750 affine_function_constant_p, affine_fn_op, affine_fn_plus,
751 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
752 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
753 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
754 (all_chrecs_equal_p): Removed.
755 * tree-data-ref.h (affine_fn, conflict_function): New types.
756 (struct subscript): Change type of conflicting_iterations_in_a
757 and conflicting_iterations_in_b.
759 2007-01-26 Steve Ellcey <sje@cup.hp.com>
762 * config/pa/pa.h (TARGET_HPUX_11): New.
763 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
764 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
766 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
767 Richard Guenther <rguenther@suse.de>
769 * tree-ssa-structalias.c (solve_graph): Handle case
770 we merged the variable to another.
772 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
774 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
777 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
778 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
780 2007-01-25 DJ Delorie <dj@redhat.com>
782 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
783 allow changes to modes which don't fit in those registers.
785 * reload1.c (choose_reload_regs): Check for invalid subregs before
786 computing their locations, not after.
788 2007-01-25 Geoffrey Keating <geoffk@apple.com>
791 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
792 save Altivec registers in an eh_return function.
793 (compute_vrsave_mask): Likewise.
794 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
795 for saving Altivec registers.
796 (rs6000_emit_prologue): Don't allocate stack twice in
797 eh_return function. Correct expected value of altivec_save_offset
798 when using save_world. Describe save of R0 to stack when using
799 save_world. Describe stack pointer adjustment when using
800 save_world. Remove duplicated eh_return parameter register saving.
801 Update sp_offset variable after save_world.
802 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
804 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
805 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
808 2007-01-25 Steve Ellcey <sje@cup.hp.com>
810 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
812 2007-01-25 Steve Ellcey <sje@cup.hp.com>
815 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
816 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
818 2007-01-25 Richard Guenther <rguenther@suse.de>
820 * doc/invoke.texi (-Wcoverage-mismatch): Document.
821 * common.opt (-Wcoverage-mismatch): New warning option.
822 * coverage.c (get_coverage_counts): Ignore coverage mismatch
823 if -Wcoverage-mismatch is given.
825 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
827 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
828 (ipcp_driver): Change to static definition.
829 Add dumping of the ifunctions.
830 (constant_val_insert): Remove unused parameter. Support for SSA.
831 (ipcp_propagate_const): Support for SSA.
832 (ipcp_profile_bb_print): Print only analyzed nodes.
833 (ipcp_replace_map_create): Remove support for Fortran constant
835 * ipa-prop.c (ipa_method_modify_stmt,
836 ipa_callsite_compute_param): Support for SSA.
837 * ipa-prop.h (ipcp_driver): Remove declaration.
838 (IS_VALID_TREE_MAP_INDEX): Add define.
840 2007-01-24 Geoffrey Keating <geoffk@apple.com>
842 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
844 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
847 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
848 constant argument to cabs and do it without checking for
849 -funsafe-math-optimizations.
851 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
853 * c-common.h (RID_FIRST_CXX0X): New.
854 (RID_LAST_CXX0X): New.
855 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
857 * c.opt (Wc++0x-compat): New.
858 * doc/invoke.texi (-Wc++0x-compat): Document.
860 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
862 * builtins.c (fold_builtin_carg): New.
863 (fold_builtin_1): Use it.
865 2007-01-24 Jan Hubicka <jh@suse.cz>
867 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
869 * ipa-inline.c (initial_insns, max_insns): Delete.
870 (compute_max_insns): New function.
871 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
872 of insns as base for code growth.
873 (cgraph_decide_inlining): Make initial_insns local; do not compute
875 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
876 * doc/invoke.texi (inline-unit-growth): Update docs.
878 2007-01-24 Jakub Jelinek <jakub@redhat.com>
880 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
881 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
882 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
883 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
884 for CPUs that have PTA_CX16 set.
887 * gimplify.c (omp_check_private): New function.
888 (gimplify_scan_omp_clauses): Use it for
889 firstprivate/lastprivate/reduction.
892 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
893 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
896 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
897 If lower_lastprivate_clauses emits some statements, append them
898 to dlist rather than body_p and to body_p append an initializer.
899 (lower_omp_for): Adjust caller.
901 2007-01-24 Steve Ellcey <sje@cup.hp.com>
903 * target.h (globalize_decl_name): New.
904 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
905 * output.h (default_globalize_decl_name): New.
906 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
908 (globalize_decl): Ditto.
909 (default_globalize_decl_name): New.
910 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
911 (ia64_handle_version_id_attribute): New.
912 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
913 (ia64_asm_output_external): Use globalize_decl_name instead
915 * doc/extend.texi (version_id): New pragma.
916 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
918 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
920 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
921 _uleb128_t and _Unwind_SWord with _sleb128_t.
922 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
923 uw_frame_state_for, uw_update_context_1): Likewise.
924 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
925 * unwind-pe.h (read_uleb128, read_sleb128,
926 read_encoded_value_with_base): Likewise.
927 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
929 2007-01-24 Richard Guenther <rguenther@suse.de>
931 * builtins.c (expand_builtin_cexpi): Get the fndecl
932 for cexp in the correct way.
934 2007-01-24 Jan Hubicka <jh@suse.cz>
936 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
938 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
940 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
941 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
942 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
943 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
944 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
945 __SIZEOF_PTRDIFF_T__.
946 * doc/cpp.texi: Documentation for the new builtin macros added.
948 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
950 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
951 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
952 instead of operands[2] to avoid access past the end of array.
954 2007-01-24 Richard Sandiford <richard@codesourcery.com>
956 * reload1.c (emit_reload_insns): Pass the reload register
957 for a non-spill output reload through forget_old_reloads_1.
959 2007-01-23 Joseph Myers <joseph@codesourcery.com>
961 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
962 _SOFT_DOUBLE if doubles use software floating-point.
963 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
964 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
965 * config/rs6000/darwin-ldouble.c: Also compile functions for
966 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
967 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
968 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
969 commented-out long double override.
970 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
971 * config/rs6000/eabispe.h: Likewise.
972 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
973 long double for non-SPE.
974 (rs6000_handle_option): Likewise.
975 (invalid_e500_subreg): Disallow more subregs involding DImode,
976 DFmode, TImode or TFmode.
977 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
979 (legitimate_lo_sum_address_p): Also check for TFmode for E500
981 (rs6000_legitimize_address): Also handle TFmode for E500 double.
982 (rs6000_legitimize_reload_address): Also handle TFmode for E500
984 (rs6000_legitimate_address): Also check for TFmode for E500
986 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
987 (spe_build_register_parallel): Handle TFmode and TCmode.
988 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
990 (function_arg): Handle TFmode and TCmode for E500 double.
991 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
993 (print_operand): Handle TFmode and TImode for %y.
994 (rs6000_generate_compare): Handle TFmode comparisons for E500
996 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
997 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
998 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
999 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
1000 TFmode for E500 double.
1001 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
1002 (floatsidf2): Enable for E500 double.
1003 (movtf_softfloat): Use rs6000_nonimmediate_operand.
1004 (extenddftf2): Change to extenddftf2_fprs.
1005 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
1006 depending on TARGET_E500_DOUBLE.
1007 (extendsftf2): Enable for E500 double.
1008 (trunctfdf2): Enable for E500 double.
1009 (trunctfsf2): Change to trunctfsf2_fprs.
1010 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
1011 depending on TARGET_E500_DOUBLE.
1012 (floatsitf2): Enable for E500 double.
1013 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
1014 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
1015 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
1016 (negtf2): Change to negtf2_internal.
1017 (negtf2): New expander.
1018 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
1019 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
1020 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
1021 (movdi_internal32): Use rs6000_nonimmediate_operand.
1022 (unnamed splitter): Likewise.
1023 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
1024 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
1025 (SPE64TF, DITI): New mode macros.
1026 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
1029 (frob_<mode>_di_2): New.
1030 (frob_tf_di_8_2): New.
1031 (frob_di_df): Change to frob_di_<mode>; allow more modes.
1033 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
1035 (frob_ti_<mode>_8_2): New.
1036 (frob_ti_tf_2): New.
1037 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
1038 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
1040 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
1041 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
1042 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
1043 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
1044 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
1045 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
1048 2007-01-23 Ian Lance Taylor <iant@google.com>
1050 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
1052 (OBJS-archive): Reformat, alphabetize.
1053 (OBJS): Change out_object_file to OBJS-md.
1055 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
1057 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
1058 JUMP_P and CALL_P predicates where applicable.
1059 * config/i386/i386.c: Ditto.
1060 * config/i386/i386.md: Ditto.
1061 * config/i386/mmx.md: Ditto.
1062 * config/i386/predicates.md: Ditto.
1064 2007-01-22 Andreas Schwab <schwab@suse.de>
1066 * config/m68k/m68k.h: Fix comment.
1068 2007-01-22 Jan Hubicka <jh@suse.cz>
1070 * passes.c (init_optimization_passes): Do not rerun
1071 pass_early_warn_uninitialized.
1073 2007-01-22 Richard Guenther <rguenther@suse.de>
1075 PR tree-optimization/30038
1076 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
1078 (execute_cse_sincos_1): Likewise.
1079 (execute_cse_sincos): Likewise.
1080 (gate_cse_sincos): Likewise.
1081 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
1082 the cexpi() canonicalization of sincos().
1083 * tree-pass.h (pass_cse_sincos): Declare.
1084 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
1086 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1088 PR rtl-optimization/29329
1089 * combine.c (replaced_rhs_insn): Rename to i2mod.
1090 (replaced_rhs_value): Rename to i2mod_new_rhs.
1091 (i2mod_old_rhs): New global variable.
1092 (combine_instructions): Adjust for above change. Save a copy of
1093 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
1094 note are substituted in the second instruction.
1095 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
1096 ditch the note if it pertains to the second eliminated register
1097 and this register is mentioned in i2mod_old_rhs.
1100 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
1102 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
1103 the insn to which the note was originally attached.
1105 2007-01-21 Jan Hubicka <jh@suse.cz>
1107 * ipa-inline.c (inlining_mode): Comment, move up.
1108 (cgraph_decide_inlining_incrementally): Do not perform inlining
1109 itself; fix handling of flattening of self recursive functions.
1110 (cgraph_find_cycles): Remove.
1111 (cgraph_flatten_node): Remove.
1112 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
1113 (try_inline): New function.
1114 (cgraph_early_inlining): Update call of
1115 cgraph_decide_inlining_incrementally. Apply inlining here.
1116 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
1118 2007-01-21 Dirk Mueller <dmueller@suse.de>
1121 * tree-vrp.c (check_array_bounds): do not warn
1122 about ADDR_EXPR's of ARRAY_REF's which are immediately
1123 used in binary expressions.
1125 2007-01-21 Ira Rosen <irar@il.ibm.com>
1127 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
1128 read_write_dep and macros for its access.
1129 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
1130 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
1131 argument, call vect_check_interleaving for every independent pair of
1132 data-refs. Mark loads that access the same memory location as a store
1134 (vect_check_dependences): Remove.
1135 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
1136 call, fix the call to vect_analyze_data_ref_dependence.
1137 (vect_analyze_data_ref_access): For statements that access the same
1138 data-ref, check that they are not stores; for loads, check that there
1139 is no store that access the same location.
1141 2007-01-20 Roger Sayle <roger@eyesopen.com>
1142 Joseph Myers <joseph@codesourcery.com>
1144 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
1145 (-fdump-rtl-stack): Likewise.
1146 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
1147 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
1148 (-mpt-fixed): Use @minus{} for minus sign.
1149 (Using Precompiled Headers): Fix under/overfull hbox.
1151 2007-01-20 Jan Hubicka <jh@suse.cz>
1153 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
1154 place within annotation.
1156 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1158 * pa.c (output_move_double): Change array size of xoperands to 4.
1160 2007-01-20 Jan Hubicka <jh@suse.cz>
1162 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
1165 2007-01-19 Ian Lance Taylor <iant@google.com>
1167 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
1168 constant to unsigned HOST_WIDE_INT before negating.
1170 2007-01-19 Ian Lance Taylor <iant@google.com>
1172 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
1174 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
1175 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1176 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
1177 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
1178 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1179 (realloc_vdef, realloc_vuse): Change second parameter to
1181 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
1183 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
1184 unsigned. Change x and lim locals to unsigned.
1185 (realloc_vdef, realloc_vuse): Change num_elem parameter to
1187 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
1188 (copy_virtual_operands): Change i and n locals to unsigned.
1190 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1193 * builtins.c (fold_builtin_1): Handle builtin fdim.
1195 2007-01-20 Jan Hubicka <jh@suse.cz>
1197 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
1198 * tree-ssa-structalias.c (compute_points_to_sets): Do call
1199 init_alias_heapvars.
1200 (init_alias_heapvars): Initialize only when not already initialized.
1201 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
1203 2007-01-19 Roger Sayle <roger@eyesopen.com>
1205 * common.opt (fsigned-zeros): New command line option.
1206 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
1207 of flag_unsafe_math_optimizations.
1208 * opts.c (set_fast_math_flags): The -ffast-math command line option
1209 implies -fno-signed-zeros.
1210 (fast_math_flags_set_p): Likewise.
1212 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
1213 the documentation of -ffast-math appropriately. Wrap long lines.
1215 2007-01-19 Steve Ellcey <sje@cup.hp.com>
1217 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
1218 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
1220 2007-01-19 Tomas Bily <tbily@suse.cz>
1222 * cgraphunit.c (cgraph_finalize_function): Updating of pid
1224 (tree_init_ic_make_global_vars): New function
1225 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
1226 (tree_gen_ic_profiler): New function
1227 (tree_gen_ic_func_profiler): New function
1228 (tree_profiling): Added calling of tree_gen_ic_func_profiler
1229 (tree_profile_hooks): Added hook for indirec/virtual calls
1230 * value-prof.c (tree_find_values_to_profile): New case for
1232 (tree_values_to_profile): Call for determining indirect/virtual
1234 (tree_indirect_call_to_profile): New function
1235 (tree_ic_transform): New function
1236 (tree_ic): New function
1237 (find_func_by_pid): New function
1238 (init_pid_map): New function
1239 (tree_value_profile_transformations): Added check for
1240 indirect/virtual call transformation
1241 * value-prof.h (enum hist_type): New counter type for
1242 indirect/virtual calls
1243 (profile_hooks): Added new hook for profiling indirect/virtual
1245 * profile.c (instrument_values): New case for indirect/virtual
1247 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
1248 (GCOV_COUNTER_V_INDIR): New counter type
1249 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
1250 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
1252 * cgraph.c: Definition of cgraph_max_pid
1253 (cgraph_create_node): Default init of pid attribute
1254 * cgraph.h: Declaration of cgraph_max_pid
1255 (struct cgraph_node): Added pid attribute
1256 * libgcov.c (__gcov_indirect_call_profiler): New function
1257 (__gcov_one_value_profiler_body): New function
1258 (__gcov_one_value_profiler): Body was moved to
1259 __gcov_one_value_profiler_body and calls it
1261 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
1263 * doc/gty.texi (Options): Document the mark_hook option to GTY.
1264 * gengtype.c (write_types_data, write_func_for_structure,
1265 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
1266 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
1267 (walk_type, write_func_for_structure): Generate the mark_hook if
1270 2007-01-19 Jan Hubicka <jh@suse.cz>
1272 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1273 'early' argument take inlining mode argument specifying whether to
1274 inline for size/speeed or all functions; add support for flattening;
1276 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1278 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1281 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1282 the name of the declared identifier.
1284 2007-01-19 Dirk Mueller <dmueller@suse.de>
1286 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1287 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1288 duplicate evaluation.
1290 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1292 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1293 Macroize expander using X87MODEF12 mode macro. Extend operand 1
1294 to XFMode, use acosxf2 and truncate result to requested mode.
1295 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1296 (asin<mode>2): Similarly, with asin expanders.
1297 (*fscalexf4_i387): Rename from *fscalexf4.
1298 (expNcorexf3): New expander.
1299 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1300 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1301 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
1302 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1303 patterns for SSE math.
1304 (exp10<mode>2): Similarly, with exp10 expanders.
1305 (exp2<mode>2): Similarly, with exp2 expanders.
1306 (expm1<mode>2): Similarly, with expm1 expanders.
1307 (ldexp<mode>3): Similarly, with ldexp expanders.
1308 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1309 Use gen_truncxf<mode>2_i387_noop to truncate result.
1311 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1313 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1314 for ColdFire targets.
1316 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
1317 Richard Sandiford <richard@codesourcery.com>
1319 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1320 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1321 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1322 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1323 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1324 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1326 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1328 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1329 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
1330 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
1331 (*adddi_dilshr32_cf): New pattern.
1332 (adddi3, subdi3): Remove first alternatives.
1334 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1336 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1337 implemented using move.l, do not use its cc result for floating-point
1340 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1342 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1343 from accepting 's' constraints if flag_pic.
1345 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1347 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1348 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1349 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1350 condition from TARGET_68881 to TARGET_HARD_FLOAT.
1352 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
1354 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1355 ColdFire alternatives.
1356 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1357 * config/m68k/m68k.md (clzsi2): Define for ColdFire
1358 architectures that support ff1 instruction.
1360 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1361 Julian Brown <julian@codesourcery.com>
1363 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1364 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1365 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
1368 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1370 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1372 (output_move_const_into_data_reg, output_move_simode_const): Delete.
1373 * config/m68k/m68k.c (const_method, const_int_cost): Take a
1374 HOST_WIDE_INT instead of an rtx.
1375 (m68k_rtx_costs): Update call accordingly.
1376 (output_move_const_into_data_reg): Likewise. Fix formatting.
1377 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1379 (output_move_simode_const): Update calls after above changes.
1380 Rework to use automatic variables and predicates like MEM_P.
1381 * config/m68k/m68k.md (pushexthisi_const): Update call to
1384 2007-01-19 Dirk Mueller <dmueller@suse.de>
1386 * tree-ssa-alias.c (perform_var_substitution): Fix typo
1389 2007-01-19 Richard Guenther <rguenther@suse.de>
1391 * builtins.c (expand_builtin_cexpi): Fall back to expanding
1392 via cexp in case sincos is not available.
1394 2007-01-19 Richard Guenther <rguenther@suse.de>
1396 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1397 * defaults.h (TARGET_HAS_SINCOS): Default to off.
1398 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1399 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1400 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1401 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1402 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1403 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1405 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1407 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1409 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1410 (atan2xf3): Directly generate RTL pattern.
1411 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1412 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1413 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1414 disable patterns for SSE math.
1415 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1416 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1417 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1418 disable patterns for SSE math.
1420 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
1422 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1424 2007-01-18 Roger Sayle <roger@eyesopen.com>
1426 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1427 VIEW_CONVERT_EXPR to the same type as it's operand.
1429 2007-01-18 David Edelsohn <edelsohn@gnu.org>
1431 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1432 configured for long double 128.
1434 2007-01-18 Mike Stump <mrs@apple.com>
1436 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1439 2007-01-18 Michael Meissner <michael.meissner@amd.com>
1441 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1444 2007-01-18 Jan Hubicka <jh@suse.cz>
1446 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1447 greater than NUM_VOP_FREE_BUCKETS.
1449 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
1451 * tree-ssa-structalias.c: Update comments.
1452 (ptabitmap_obstack): Removed.
1454 (oldpta_obstack): Ditto.
1455 (stats): Add a few members.
1456 (struct variable_info): Remove node, complex, address_taken, and
1457 indirect_target members. Add oldsolution member.
1458 (new_var_info): Do not initialize removed members.
1459 (constraint_expr_type): Remove INCLUDES.
1460 (constraint_graph): Add size, implicit_preds, rep,
1461 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1462 (FIRST_REF_NODE): New macro.
1463 (LAST_REF_NODE): Ditto.
1464 (FIRST_ADDR_NODE): Ditto.
1465 (find): New function.
1467 (dump_constraint): Do not handle INCLUDES.
1468 (insert_into_complex): Do not insert duplicate constraints.
1469 (condense_varmap_nodes): Renamed and rewritten into ...
1470 (merge_node_constraints): This. Also fix bug in handling of
1471 offseted copy constraints.
1472 (clear_edges_for_node): No longer need to deal with preds at all,
1473 or removing associated preds/succs.
1474 (merge_graph_nodes): Deal with indirect_cycles.
1475 Don't deal with predecessors.
1476 (add_implicit_graph_edge): New function.
1477 (add_pred_graph_edge): Ditto.
1478 (add_graph_edge): Don't deal with predecessors.
1479 (build_constraint_graph): Removed.
1480 (build_pred_graph): New function.
1481 (build_succ_graph): Ditto.
1482 (struct scc_info): Removed in_component. Added roots, dfs, and
1483 node_mapping. Remove visited_index, unification_queue.
1484 (scc_visit): Deal with union-find we do now.
1485 Deal with cycles with REF nodes.
1486 (collapse_nodes): Renamed and rewritten to ...
1487 (unify_nodes): This.
1488 (process_unification_queue): Removed.
1489 (topo_visit): Cleanup
1490 (do_da_constraint): Use find.
1491 (do_sd_constraint): Ditto.
1492 (do_ds_constraint): Ditto.
1493 (do_complex_constraint): Ditto.
1494 (init_scc_info): Update for removed and added members.
1495 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1496 (find_indirect_cycles): This.
1497 (equivalence_class): New variable.
1498 (label_visit): New function.
1499 (perform_variable_substitution): Rewritten.
1500 (free_var_substitution_info): New function.
1501 (find_equivalent_node): Ditto.
1502 (move_complex_constraints): Ditto.
1503 (eliminate_indirect_cycles): Ditto.
1504 (solve_graph): Only propagate changed bits.
1505 Use indirect cycle elimination.
1507 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1508 (tree_id_eq): Renamed to ...
1509 (tree_vi_eq): This. Update for member change
1510 (insert_id_for_tree): Renamed and rewritten to ...
1511 (insert_vi_for_tree): This.
1512 (lookup_id_for_tree): Renamed and rewritten to ...
1513 (lookup_vi_for_tree): This.
1514 (get_id_for_tree): Renamed and rewritten to ...
1515 (get_vi_for_tree): Ditto.
1516 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1517 (process_constraint): Don't handle INCLUDES.
1518 Remove special ADDRESSOF case.
1519 (find_func_aliases): Rewrite to use vi functions instead of id
1521 (create_function_info_for): Ditto.
1522 (create_variable_info_for): Ditto.
1523 (intra_create_variable_infos): Ditto.
1524 (merge_smts_into): Ditto.
1525 (find_what_p_points_to): Ditto.
1526 (init_base_vars): Ditto.
1527 (init_alias_vars): Ditto.
1528 (remove_preds_and_fake_succs): New function.
1529 (dump_sa_points_to_info): Dump new stats.
1530 (dump_solution_for_var): Use find.
1531 (set_used_smts): Fix formatting.
1532 (compute_points_to_sets): Updated for new functions.
1533 (ipa_pta_execute): Ditto.
1535 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1536 Richard Sandiford <richard@codesourcery.com>
1538 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1540 * calls.c (expand_call): If the return value is a PARALLEL,
1541 extract its first member.
1542 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1543 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1544 use FUNCTION_EXTRA_EPILOGUE.
1545 (m68k_function_value): Return a PARALLEL if the return value
1546 is of a pointer type.
1547 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1548 (FUNCTION_EXTRA_EPILOGUE): Remove.
1549 * config/m68k/m68k.md (D0_REG): New constant.
1551 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1553 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1554 output a NOP for empty epilogues.
1556 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1558 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1559 before function. Extend register save check to include all
1560 registers, not just integer ones.
1562 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1564 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1565 (m68k_use_return_insn): ...this.
1566 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1567 * config/m68k/m68k.c (use_return_insn): Rename to...
1568 (m68k_use_return_insn): ...this.
1569 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1572 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1574 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1575 nearest even, fix denormal rounding overflow.
1577 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1579 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1580 f%$smove and f%$move.
1581 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1582 (extendsfdf2_cf): Use fdmove instead of f%&move.
1583 (truncdfsf2_cf): Use fsmove instead of f%$smove.
1584 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1586 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1588 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1591 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1593 * real.h (real_format): Add a canonical_nan_lsbs_set field.
1594 (coldfire_single_format): Declare.
1595 (coldfire_double_format): Likewise.
1596 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1597 of qnan_msb_set to determine the lower bits of a canonical
1599 (encode_ieee_double): Likewise.
1600 (encode_ieee_quad): Likewise.
1601 (ieee_single_format): Initialize canonical_nan_lsbs_set.
1602 (mips_single_format): Likewise.
1603 (ieee_double_format): Likewise.
1604 (mips_double_format): Likewise.
1605 (ieee_extended_motorola_format): Likewise.
1606 (ieee_extended_intel_96_format): Likewise.
1607 (ieee_extended_intel_128_format): Likewise.
1608 (ieee_extended_intel_96_round_53_format): Likewise.
1609 (ibm_extended_format): Likewise.
1610 (mips_extended_format): Likewise.
1611 (ieee_quad_format): Likewise.
1612 (mips_quad_format): Likewise.
1613 (vax_f_format): Likewise.
1614 (vax_d_format): Likewise.
1615 (vax_g_format): Likewise.
1616 (i370_single_format): Likewise.
1617 (i370_double_format): Likewise.
1618 (decimal_single_format): Likewise.
1619 (decimal_double_format): Likewise.
1620 (decimal_quad_format): Likewise.
1621 (c4x_single_format): Likewise.
1622 (c4x_extended_format): Likewise.
1623 (real_internal_format): Likewise.
1624 (coldfire_single_format): New real_format.
1625 (coldfire_double_format): Likewise.
1626 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1627 canonical_nan_lsbs_set.
1628 (pdp11_d_format): Likewise.
1629 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1630 if TARGET_COLDFIRE_CPU.
1632 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1634 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1635 (output_call, m68k_legitimize_call_address): Declare.
1636 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1637 parenthesees. Add support for a 'W' constraint.
1638 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1639 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1640 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1641 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1643 (override_options): Initialize them. Do not set flag_no_function_cse
1644 for TARGET_ID_SHARED_LIBRARY.
1645 (m68k_output_pic_call): Delete.
1646 (m68k_legitimize_call_address): New function.
1647 (print_operand): Remove the %o prefix. Handle the %p prefix.
1648 (output_call): New function.
1649 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
1650 target address from the GOT if symbolic jumps are not allowed.
1651 * config/m68k/m68k.md (call, general_operand): Do not set
1652 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
1653 Merge separate flag_pic and !flag_pic define_insns into...
1654 (*call, *call_value): ...these new patterns. Match the address
1655 rather than the containing MEM and require it to be a call_operand.
1656 Use output_call to generate the asm template.
1657 * config/m68k/predicates.md (const_call_operand): New predicate.
1658 (call_operand): Likewise.
1660 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1662 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1663 (STACK_GROWS_DOWNWARD): Define to 1.
1664 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1665 comments, emphasizing that these values are only defaults.
1666 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1667 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1669 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1671 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1673 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1674 Richard Sandiford <richard@codesourcery.com>
1676 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1677 (FP_REGNO_P): New macros.
1678 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1679 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1680 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1681 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1683 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1685 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1686 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1688 * config/m68k/t-floatlib: New file, extracting common code from...
1689 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1690 * config/m68k/t-uclinux: Here.
1691 * config/m68k/fpgnulib.c: Do not compile extendeed precision
1692 routines on ColdFire targets.
1694 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1696 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1697 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1698 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1699 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1700 explicitly set MOTOROLA to 1.
1701 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1703 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1704 Nathan Sidwell <nathan@codesourcery.com>
1706 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1707 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1708 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1709 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1710 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1711 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1712 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1713 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1714 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1715 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1716 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1717 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1718 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1719 (DBX_REGISTER_NUMBER): Delete.
1720 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1721 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1722 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1724 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1726 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1727 * config/flat.h: New file.
1728 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1730 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1731 indirect references for -msep-data or -mid-shared-library.
1732 Do not use PC-relative code addresses either.
1733 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1735 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1736 for shared libraries and crt1.o for executables. Use crti.o and
1738 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1739 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1740 Do not add -elf2flt or -shared-lib-id options here.
1741 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1742 Pass -shared-lib-id if -mid-shared-library, taking the library
1743 identifier from -mshared-library-id if given, otherwise
1745 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1746 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1747 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1748 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1749 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1750 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1752 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1753 jump sequence for ISA A and ISA A+.
1754 (PICJUMP): Likewise.
1756 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1758 * config.gcc (m68k-*-uclinux*): Base the port on the common
1759 and m68k GNU/Linux files rather than on the generic ELF ones.
1760 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1761 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1763 2007-01-18 Julian Brown <julian@codesourcery.com>
1764 Richard Sandiford <richard@codesourcery.com>
1766 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1767 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1768 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1770 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1772 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1774 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1776 * doc/install.texi: Document m68k-uclinuxoldabi.
1777 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1778 * config/m68k/uclinux-oldabi.h: New file, copied from
1779 config/m68k/uclinux.h.
1781 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1783 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1784 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1785 (m32r_compute_frame_size): Use unsigned for regno.
1786 (m32r_reload_lr): Use gen_frame_mem.
1788 (m32r_output_function_epilogue): Don't output the function epilogue
1790 (m32r_expand_epilogue): New.
1791 (direct_return): Return false if the function has the interrupt
1793 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1794 * config/m32r/m32r.md (epilogue): New expander.
1795 (return_lr, return_rte): New insns.
1796 (return): Make it expander.
1797 (return_normal): New expander.
1799 2007-01-18 Josh Conner <jconner@apple.com>
1802 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1803 support for UNLE, UNLT, UNGE, and UNGT.
1805 2007-01-18 Jan Hubicka <jh@suse.cz>
1807 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1808 * tree-tailcall (add_virtual_phis): Likewise.
1809 (optimize_tail_call): Return TODOs.
1810 (execute_tail_calls): Return TODOs.
1811 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1812 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1814 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1815 Cleanup cfg using TODO.
1816 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1817 * passes.c (execute_function_todo): When cleanup did something, remove
1819 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1820 (make_edges): Don't cleanup_cfg.
1822 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1824 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1825 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1826 operate on the top of stack.
1827 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1828 register handling of instructions that output to the second
1830 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1831 (move_for_stack_reg): Special-case check for dead destination
1832 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1834 * config/i386/i386.md (UNSPEC_TAN): New constant.
1835 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1836 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1837 to correctly model move of constant 1.0 to top stack slot.
1838 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1839 (unnamed peephole2 pattern): Remove corresponding peephole2
1840 pattern that optimizes tan insn and loading of constant 1.0.
1841 (tanxf2): Use fptanxf4_i387.
1842 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1843 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1844 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1845 disable patterns for SSE math.
1846 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1847 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1849 2007-01-18 Dirk Mueller <dmueller@suse.de>
1850 Richard Guenther <rguenther@suse.de>
1853 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1854 * common.opt (Warray-bounds): Add new warning option.
1855 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1857 * Makefile.in: make tree-vrp.o depend on toplev.h
1858 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1860 (check_array_refs, check_array_bounds, check_array_ref): New.
1862 2007-01-18 Jan Hubicka <jh@suse.cz>
1864 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1865 (execute_ssa_ccp): Return flags conditionally.
1866 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1868 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1870 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1872 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1873 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1875 * Makefile.c (GTFILES): Add cfglayout.h.
1876 * gengtype.c (open_base_files): Likewise.
1877 * cfglayout.c (cfg_layout_function_footer,
1878 cfg_layout_function_header) Reindent to make gengtype happy.
1879 * cfglayout.h (cfg_layout_function_footer,
1880 cfg_layout_function_header): Add GTY(()) marker.
1882 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1884 2007-01-18 Ben Elliston <bje@au.ibm.com>
1886 * genautomata.c (write_automata): Include xstrerror output in the
1887 error message if writing the DFA description file fails.
1889 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1891 * config/mips/mips-protos.h (mips_output_external): Make it
1893 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1894 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1895 (extern_list): Likewise.
1896 (extern_head): Likewise.
1897 (TARGET_ASM_FILE_END): Likewise.
1898 (mips_file_end): Likewise.
1899 (mips_output_external): Rewritten.
1901 2007-01-18 Ben Elliston <bje@au.ibm.com>
1903 * genpreds.c (write_insn_preds_c): Only write out the function
1904 body for regclass_for_constraint if we have register constraints.
1906 2007-01-17 Tom Tromey <tromey@redhat.com>
1908 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1909 * doc/install.texi (Testing): Use sourceware.org.
1910 (Binaries): Likewise.
1911 (Specific): Likewise.
1912 * doc/contrib.texi (Contributors): Use sourceware.org.
1914 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1916 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1917 (AVR_ENHANCED): Rename to ...
1918 (AVR_HAVE_MUL): ... new.
1919 (avr_enhanced_p): Rename to ...
1920 (avr_have_mul_p): ... new.
1921 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1922 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1923 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1924 (avr_have_mul_p): ... new.
1925 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1926 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1927 'avr_enhanced_p' and 'enhanced'.
1928 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1929 AVR_HAVE_MUL instead of AVR_ENHANCED.
1930 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1931 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1932 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1933 instead of AVR_ENHANCED.
1934 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1935 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1936 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1938 2007-01-17 Ian Lance Taylor <iant@google.com>
1940 * vec.h (VEC_reserve_exact): Define.
1941 (vec_gc_p_reserve_exact): Declare.
1942 (vec_gc_o_reserve_exact): Declare.
1943 (vec_heap_p_reserve_exact): Declare.
1944 (vec_heap_o_reserve_exact): Declare.
1945 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1947 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1949 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1950 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1951 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1952 * vec.c (calculate_allocation): Add exact parameter. Change all
1954 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1955 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1956 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1957 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1958 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1959 (vec_heap_p_reserve_exact): New function.
1960 (vec_heap_o_reserve_exact): New function.
1962 2007-01-17 Jan Hubicka <jh@suse.cz>
1964 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1966 2007-01-17 Eric Christopher <echristo@apple.com>
1968 * config.gcc: Support core2 processor.
1970 2007-01-16 Jan Hubicka <jh@suse.cz>
1972 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1973 the names, just unlink the chain so we don't crash on dangling pointers
1976 2007-01-16 Jan Hubicka <jh@suse.cz>
1978 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1979 * tree-pass.h: Reorder to make IPA passes appear toegher.
1980 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1981 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1982 parameters, do not call early inliner.
1983 * ipa-inline.c: Update comments. Include tree-flow.h
1984 (cgraph_decide_inlining): Do not compute inlining parameters.
1985 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1986 be called with function context set up.
1987 (pass_ipa_inline): Remove unreachable functions before pass.
1988 (cgraph_early_inlining): Simplify assuming to be called from the
1990 (pass_early_inline): New pass.
1991 (cgraph_gate_ipa_early_inlining): New gate.
1992 (pass_ipa_early_inline): Turn into simple wrapper.
1993 (compute_inline_parameters): New function.
1994 (gate_inline_passes): New gate.
1995 (pass_inline_parameters): New pass.
1996 (apply_inline): Move here from tree-optimize.c
1997 (pass_apply_inline): New pass.
1998 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
2000 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
2002 (tree_function_versioning): Do not allocate dummy struct function.
2003 * tree-inline.h (optimize_inline_calls): Update prototype.
2004 * tree-optimize.c (execute_fixup_cfg): Export.
2005 (pass_fixup_cfg): Remove
2006 (tree_rest_of_compilation): Do not apply inlines.
2007 * tree-flow.h (execute_fixup_cfg): Declare.
2008 * Makefile.in (gt-passes.c): New.
2009 * passes.c: Include gt-passes.h
2010 (init_optimization_passes): New passes.
2011 (nnodes, order): New static vars.
2012 (do_per_function_toporder): New function.
2013 (execute_one_pass): Dump current pass here.
2014 (execute_ipa_pass_list): Don't dump current pass here.
2016 2007-01-16 Janis Johnson <janis187@us.ibm.com>
2018 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
2019 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
2021 2007-01-16 David Edelsohn <edelsohn@gnu.org>
2023 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
2024 (strong_alias): Define.
2025 (__gcc_qmul): Provide non-FMA for soft-float.
2031 (__gcc_qunord): New.
2041 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
2042 soft-float functions.
2043 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
2044 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
2045 about long double soft float.
2047 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
2048 Tehila Meyzels <tehila@il.ibm.com>
2050 * tree-vectorizer.h (is_pattern_stmt_p): New.
2051 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
2052 formatting (tabs instead of spaces). Cleanup and clarify setting
2053 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
2054 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
2055 (vectorizable_type_demotion): Check that types are integral.
2056 (vectorizable_type_promotion): Likewise.
2057 (vectorizable_store): Fix typo. Eliminate new-line at end of
2060 2007-01-16 Jan Hubicka <jh@suse.cz>
2062 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
2065 2007-01-15 Eric Christopher <echristo@apple.com>
2067 * ifcvt.c: Include vec.h, vecprim.h.
2068 (check_cond_move_block): New argument regs.
2069 Reorganize. Add registers used to regs.
2070 (cond_move_process_if_block): Use regs set above as
2073 2007-01-15 Eric Christopher <echristo@apple.com>
2075 * config/darwin.h: Update copyright.
2076 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
2077 (LINK_COMMAND_SPEC): Add -u.
2078 (LINK_SPEC): Fix umbrella for above.
2080 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
2082 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
2085 2007-01-15 Tom Tromey <tromey@redhat.com>
2087 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
2088 * doc/install.texi (Testing): Don't mention jacks.
2089 (Configuration): Document --enable-java-maintainer-mode. Move
2090 --with-java-home to libgcj-specific section. Document
2092 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
2094 2007-01-15 Jan Hubicka <jh@suse.cz>
2096 * tree-ssa-dce.c (DCE_TODOs): New.
2097 (propagate_necessity): Return if something changed.
2098 (eliminate_unnecessary_stmts): Likewise.
2099 (perform_tree_ssa_dce): Return TODO flags when needed.
2100 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
2102 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
2104 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
2105 (fyl2x_extend<mode>xf3_i387): New insn pattern.
2106 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
2107 patterns using X87MODEF12 mode macro. Extend operand 1
2108 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2109 (log10<mode>2): Ditto.
2110 (log2<mode>2): Ditto.
2111 (log1p<mode>2): Ditto.
2112 (logb<mode>2): Ditto.
2113 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
2114 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
2115 (*fxtractxf3_i387): Rename from *fxtractxf3.
2116 (fxtract_extend<mode>xf3_i387): New insn pattern.
2117 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
2118 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
2119 and gen_fyl2xxf3_i387().
2121 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
2123 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
2124 * cfgloopmanip.c (fix_loop_placement): Made static. Use
2125 get_loop_exit_edges. Changed return type to bool.
2126 * cfgloop.h (fix_loop_placement): Declaration removed.
2128 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
2130 * param.h (MIN_VECT_LOOP_BOUND): New.
2131 * params.def (MIN_VECT_LOOP_BOUND): New.
2132 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
2133 argument - minimum threshold for number of iterations.
2134 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
2135 argument to declaration.
2136 * tree-vect-analyze.c (vect_analyze_operations): Check value of
2137 MIN_VECT_LOOP_BOUND.
2138 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
2139 slpeel_tree_peel_loop_to_edge with additional argument.
2140 (vect_do_peeling_for_alignment): Likewise.
2141 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
2143 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
2146 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
2149 2007-01-14 Jan Hubicka <jh@suse.cz>
2151 * tree-dfa.c (remove_referenced_var): New function.
2152 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
2153 prune referenced vars list too.
2154 * tree-flow.h (remove_referenced_var): Declare.
2156 2007-01-14 Jan Hubicka <jh@suse.cz>
2158 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
2160 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
2161 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
2162 * except.h (verify_eh_throw_table_statements): Kill prototype.
2163 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
2164 into call exrepssions.
2165 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
2167 * tree-cfg.c: Include pointer-set.h.
2168 (verify_node_sharing): Work on pointer set.
2169 (verify_eh_throw_stmt_node): New.
2170 (verify_stmts): Use pointers sets, verify throw_stmt.
2172 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
2174 * ipa-reference.c (analyze_function): Consider also addresses taken
2177 2007-01-12 Roger Sayle <roger@eyesopen.com>
2179 * c-typeck.c (null_pointer_constant_p): Replace use of
2180 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2181 (build_c_cast): Likewise.
2183 2007-01-12 Roger Sayle <roger@eyesopen.com>
2185 * tree.h (force_fit_type_double): Remove unused final argument.
2186 * c-common.c (constant_expression_warning): Replace use of
2187 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2188 (convert_and_check): Likewise.
2189 (shorten_compare): Update call to force_fit_type_double.
2190 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
2191 * convert.c (convert_to_pointer): Update call to
2192 force_fit_type_double.
2193 * fold-const.c (force_fit_type_double): Remove overflowed_const
2195 (int_const_binop, fold_convert_const_int_from_int,
2196 fold_convert_const_int_from_real, fold_div_compare,
2197 fold_sign_changed_comparison, fold_unary, fold_negate_const,
2198 fold_abs_const, fold_not_const): Remove the final argument from
2199 calls to force_fit_type_double.
2201 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
2203 * configure.ac: Set insn to "nop" for spu-*-* also.
2204 * configure: Regenerate.
2206 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
2208 * builtins.def : Add BUILT_IN_FREE.
2210 2007-01-12 Jan Hubicka <jh@suse.cz>
2212 PR tree-optimization/30443
2213 * tree-inline.c (tree_function_versioning): Do not optimize when
2214 cloning for inlining.
2216 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
2218 * doc/loop.texi: Document recording of loop exits.
2219 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
2220 (update_single_exits_after_duplication,
2221 update_single_exit_for_duplicated_loop,
2222 update_single_exit_for_duplicated_loops): Removed.
2223 (duplicate_loop_to_header_edge): Do not call
2224 update_single_exits_after_duplication and
2225 update_single_exit_for_duplicated_loops.
2226 (loop_version): Do not update single_exit information.
2227 (fix_loop_structure): Use record_loop_exits instead of
2228 mark_single_exit_loops.
2229 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
2230 the lists of loop exits.
2231 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
2232 split_edge, merge_blocks): Update the lists of loop exits.
2233 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
2234 loop_optimizer_init.
2235 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
2236 of mark_single_exit_loops.
2237 (loop_optimizer_finalize): Call release_recorded_exits.
2238 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
2239 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
2240 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
2241 update single exit information.
2242 * lambda-code.c (perfect_nestify): Ditto.
2243 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
2244 (mark_single_exit_loops): Removed.
2245 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
2246 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
2247 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
2249 (get_loop_exit_edges, single_exit): Use recorded exit lists.
2250 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
2251 (verify_loop_structure): Verify consistency of the exit lists.
2252 (flow_loops_find): Use alloc_loop. Initialize exits hash.
2253 (set_single_exit): Removed.
2254 * cfgloop.h (struct loop_exit): New function.
2255 (struct loop): single_exit_ field replaced by exits field.
2256 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
2257 (struct loops): Added exits hash.
2258 (mark_single_exit_loops, set_single_exit): Declaration removed.
2259 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
2261 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2263 * doc/invoke.texi: Avoid use of @headitem.
2265 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2267 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
2269 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2271 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2274 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2275 Richard Sandiford <richard@codesourcery.com>
2277 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2278 * config/m68k/m68k.opt: Resort options.
2279 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2281 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2282 Richard Sandiford <richard@codesourcery.com>
2284 * doc/invoke.texi: Document the macros that are defined by
2285 m68k's -mtune and -mhard-float options.
2286 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2287 (m68k_cpp_cpu_family): Likewise.
2288 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2289 of __ucfv*__ macros. Define __mcffpu__ if generating code for
2290 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
2291 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2292 (m68k_cpp_cpu_family): Likewise.
2294 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2296 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2297 macros besides mc68000 as tuning macros. Use a switch statement
2298 to set them and mcpu32.
2300 2007-01-12 Julian Brown <julian@codesourcery.com>
2302 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2303 (TARGET_68040_ONLY): Rename to...
2304 (TARGET_68040): ...this.
2305 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2306 * config/m68k/m68k.md: Likewise.
2308 2007-01-12 Julian Brown <julian@codesourcery.com>
2309 Nathan Sidwell <nathan@codesourcery.com>
2310 Richard Sandiford <richard@codesourcery.com>
2312 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2313 the -mcpu= argument associated with the --with-cpu setting.
2314 Define M68K_DEFAULT_TUNE to the default -mtune= option,
2315 if different from the one implied by the -mcpu setting.
2316 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2317 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
2318 * doc/invoke.texi: Mention ColdFire in the introduction to the
2319 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
2320 -mno-div and -mhard-float options. Make -m68881 a synonym for
2321 -mhard-float. Document the previously-undocumented -m5206e,
2322 -m528x, -m5307 and -m5407 options. Tweak the existing option
2323 documentation for consistency.
2324 * doc/install.texi: Mention new --with-cpu arguments.
2325 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2326 default CPU if neither -mcpu nor -march are specified.
2327 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2328 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2329 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
2330 using m68k_tune to decide between families that implement the
2331 same ISA. Use m68k_tune to set __mcfv4e__.
2332 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2333 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2334 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2335 (FL_ISA_C, FL_ISA_MMU): New macros.
2336 (MASK_COLDFIRE): Delete.
2337 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2338 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2339 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2340 (TARGET_HARD_FLOAT): Do not define here.
2341 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2342 (TUNE_68000): New macro.
2343 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2344 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2345 (TUNE_CFV2): Redefine in terms of m68k_tune.
2346 (uarch_type, target_device, fpu_type): New enums.
2347 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2348 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2349 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2350 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2351 (FL_FOR_isa_c): New macros.
2352 (m68k_isa): New enum.
2353 (m68k_target_selection): New structure.
2354 (all_devices, all_isas, all_microarchs): New tables.
2355 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2356 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2357 (MASK_ALL_CPU_BITS): Delete.
2358 (m68k_find_selection): New function.
2359 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2360 Map the legacy target options to a combination of the new ones.
2361 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2362 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
2363 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2365 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2366 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2368 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2369 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2370 * config/m68k/m68k-devices.def: New file.
2372 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2373 Nathan Sidwell <nathan@codesourcery.com>
2375 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2376 (SUBTARGET_EXTRA_SPECS): New macros.
2377 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2378 use %(asm_cpu_spec) instead.
2379 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2380 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2381 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2382 (EXTRA_SPECS): Rename to...
2383 (SUBTARGET_EXTRA_SPECS): ...this.
2385 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2386 Richard Sandiford <richard@codesourcery.com>
2387 Julian Brown <julian@codesourcery.com>
2389 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2390 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2391 to the configuration's default CPU.
2392 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2393 Remove default masks.
2394 (m680[012]0-*-*): Set the default with_cpu to the first part of
2396 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2397 (m68k*-*-linux): Extend the --with-cpu handling to...
2398 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
2399 and m68010. Don't set target_cpu_default2.
2400 * doc/install.texi: Document --with-cpu for m68k.
2401 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2402 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2403 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2404 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2405 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2406 (ASM_SPEC): Remove use of %(asm_cpu_default).
2407 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2408 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2409 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
2410 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2411 __HAVE_FPU__ if TARGET_HARD_FLOAT.
2412 (TARGET_DEFAULT): Delete.
2413 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2414 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2415 (CPP_CPU_SPEC): Delete.
2416 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2417 the appropriate string.
2418 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2419 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2420 (ASM_SPEC): Don't use %(asm_default_spec).
2421 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2422 TARGET_DEFAULT and add MASK_68881.
2423 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2425 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2427 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2428 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2429 MASK_68010 alongside MASK_68020.
2430 * doc/invoke.texi: Document -m68010.
2431 * config/m68k/m68k.opt (m68010): New.
2432 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2434 (TUNE_68010): New macro.
2435 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2436 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2437 (M68K_CPU_m68332): Add MASK_68010.
2438 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2439 fallback definition.
2440 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2442 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2443 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
2444 to all entries that use MASK_68020.
2445 (output_move_simode_const, output_move_himode, output_move_qimode)
2446 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2447 instead of TARGET_68020 to select clr behavior. Remove comment
2448 about there being no TARGET_68010.
2449 * config/m68k/m68k.md: Likewise throughout.
2451 2007-01-12 Julian Brown <julian@codesourcery.com>
2453 * config/m68k/m68k.h (TARGET_ISAB): New macro.
2454 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2455 * config/m68k/m68k.md: Likewise.
2457 2007-01-12 Julian Brown <julian@codesourcery.com>
2459 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2460 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2462 2007-01-12 Julian Brown <julian@codesourcery.com>
2464 * config/m68k/m68k.h (TUNE_68040_60): New macro.
2465 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2466 * config/m68k/m68k.md: Likewise.
2468 2007-01-12 Julian Brown <julian@codesourcery.com>
2469 Richard Sandiford <richard@codesourcery.com>
2471 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2472 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2473 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2475 (TARGET_CPU32): Rename to...
2476 (TUNE_CPU32): ...this.
2477 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2478 (TUNE_CFV2): New macros.
2479 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2480 remove conditions that are implied by TARGET_68020.
2481 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2482 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2483 (m68k_output_function_epilogue): Likewise.
2484 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
2485 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
2486 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2487 non-68000 timings. Refactor multiplication and division costs.
2488 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2489 TUNE_CPU32 instead of TARGET_CPU32.
2490 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2491 and TUNE_68060 instead of TARGET_68060.
2492 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2493 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2495 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2496 && !TARGET_COLDFIRE" to choose between moveq and clr.
2497 Likewise in the unnamed movsf pattern.
2498 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2499 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
2502 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2504 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2505 of tabbing before backslashes.
2507 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2509 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2510 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2511 (ASM_OUTPUT_INTERNAL_LABEL): Define.
2513 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2515 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2516 find_interesting_uses_cond.
2517 (find_interesting_uses_cond): Use extract_cond_operands.
2518 (rewrite_use_compare): Use extract_cond_operands and
2519 force_gimple_operand_bsi. Do not call update_stmt.
2520 (determine_use_iv_cost_condition): Use extract_cond_operands.
2521 Return cheaper of using original bound and changing the exit bound.
2523 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2525 PR tree-optimization/29516
2526 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2527 most_expensive_mult_to_index, addr_to_parts,
2528 create_mem_ref, maybe_fold_tmr): Make the type of
2529 fields of TARGET_MEM_REF sizetype.
2530 (move_fixed_address_to_symbol, move_pointer_to_base):
2532 * tree.def (TARGET_MEM_REF): Add comment on types of
2535 2007-01-11 Joseph Myers <joseph@codesourcery.com>
2537 * c-common.c (vector_types_convertible_p): Treat opaque types as
2538 always convertible if they have the same size, but not otherwise.
2540 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
2542 * ifcvt.c (struct noce_if_info): Add comments to the fields.
2543 Remove the b_unconditional field.
2544 (noce_try_sign_mask): Do not look at b_unconditional.
2545 (noce_process_if_block): Do not use merge_if_blocks. Update
2546 the CFG here. Do not set b_unconditional.
2547 (cond_move_process_if_block): Likewise.
2548 (find_cond_trap): Likewise.
2549 (check_cond_move_block): Require simple jump insns at the end
2552 2007-01-11 Jan Hubicka <jh@suse.cz>
2554 PR tree-optimization/1046
2555 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2556 when alias info is not ready.
2557 (pass_tail_recursion): Do not require aliasing.
2558 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2559 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2560 * tree-ssa-copy.c (pass_copy_prop): Likewise.
2561 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2562 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2563 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2564 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2566 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2567 When aliasing is not build, mark statement as volatile.
2569 2007-01-11 Tom Tromey <tromey@redhat.com>
2571 PR preprocessor/15185, PR preprocessor/20989:
2572 * doc/cppopts.texi <-MT>: Update description of algorithm for
2573 computing default target.
2574 <-M, -MD>: Reword "basename" text.
2576 2007-01-11 Roger Sayle <roger@eyesopen.com>
2578 * builtins.c (expand_builtin_pow, expand_builtin_powi,
2579 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2580 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2581 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2582 fold_builtin_bswap, real_constp, fold_builtin_pow,
2583 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2584 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2585 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2586 * convert.c (convert_to_pointer): Likewise.
2587 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2588 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2589 const_binop, fold_convert_const_int_from_int,
2590 fold_convert_const_int_from_real,
2591 fold_convert_const_real_from_real, sign_bit_p,
2592 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2593 fold_sign_changed_comparison, fold_unary, fold_comparison,
2594 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2595 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2596 * print-tree.c (print_node_brief, print_node): Likewise.
2597 * stor-layout.c (place_field, layout_type): Likewise.
2598 * tree-chrec.c (keep_cast): Likewise.
2599 * tree.c (build_vector, build_real, build_real_from_int_cst,
2600 build_complex): Likewise.
2602 2007-01-11 Roger Sayle <roger@eyesopen.com>
2604 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
2605 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2607 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
2609 * configure.ac (strict1_warn): Rename to strict_warn.
2610 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2611 cc_set_by_configure, quoted_cc_set_by_configure,
2612 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2613 all_boot_languages, all_stagestuff): Remove.
2614 (target_list): Remove bootstrap targets.
2615 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2616 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2617 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2618 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2619 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2620 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2621 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2622 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2623 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2624 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2625 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2626 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2627 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2628 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2629 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2630 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2631 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2632 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2633 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2634 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2635 (mostlyclean): Adjust.
2636 (clean, distclean): Don't mention bootstrap stuff.
2637 * configure: Regenerate.
2638 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2639 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2640 treelang/config-lang.in (stagestuff): Remove.
2641 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2642 stageprofile, stagefeedback, stagestuff): Remove mention.
2644 2007-01-11 Nick Clifton <nickc@redhat.com>
2646 * config/mcore/predicates.md (mcore_general_movesrc_operand):
2648 (mcore_general_movdst_operand): Do not accept CONST_INTs.
2649 (mcore_arith_K_S_operand): Run the test for the S constraint not
2650 the test for the M constraint.
2651 (mcore_addsub_operand): Do not accept integer values that are
2652 larger than 32 bits.
2653 * config/mcore/mcore.md: Remove unused constraints from split.
2654 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2656 (allocate_stack): Likewise.
2657 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2658 operands to 32 bits.
2659 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2661 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2662 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2663 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2664 mcore_output_bclri, mcore_output_andn, output_inline_const,
2665 mcore_output_move, mcore_output_movedouble): Likewise.
2666 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2667 (output_inline_const): Likewise.
2668 (output_inline_const): Fix format strings used in sprintf
2670 * config/mcore/mcore-protos.h: Update prototypes for changed
2671 functions in mcore.c.
2672 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2673 HOST_WIDE_INT and not int.
2674 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2675 CONST_OK_FOR_N): Likewise.
2676 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2677 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2680 2007-01-10 Jan Hubicka <jh@suse.cz>
2682 * tree-vrp.c (remove_range_assertions): Release defs.
2683 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2684 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2686 2007-01-10 Paul Brook <paul@codesourcery.com>
2688 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2689 * config/arm/arm.md (mulsi3subsi): New insn.
2691 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
2693 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2694 (tree_transform_and_unroll_loop): New.
2695 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2698 2007-01-10 Robert Kennedy <jimbob@google.com>
2700 * fold-const.c (fold_comparison): Fold comparisons like (x *
2701 1000 < 0) to (x < 0).
2703 2007-01-10 Ian Lance Taylor <iant@google.com>
2705 * tree-pretty-print.c (dump_generic_node): Print parentheses when
2706 operands have the same priority.
2708 2007-01-10 Tom Tromey <tromey@redhat.com>
2710 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2711 (fold_binary): Likewise.
2712 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2713 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2714 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2715 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2716 * langhooks.h (struct lang_hooks): Removed field
2717 'can_use_bit_fields_p'.
2719 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
2721 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2723 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
2725 * function.c (get_last_funcdef_no): New function.
2726 * function.h (get_last_funcdef_no): Declare.
2727 * tree-inline.c (initialize_cfun): Add initialization.
2728 (tree_function_versioning): Cleanup.
2730 2007-01-10 Jan Hubicka <jh@suse.cz>
2732 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2735 2007-01-10 Sa Liu <saliu@de.ibm.com>
2736 Ben Elliston <bje@au.ibm.com>
2738 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2739 for SAVE_NONLOCAL and Pmode for any other save level.
2740 * spu-protos.h (spu_restore_stack_block): Declare.
2741 * spu.md (save_stack_block): Remove.
2742 (restore_stack_block): Call spu_restore_stack_block.
2743 * spu.c (spu_restore_stack_block): New function.
2744 (spu_expand_epilogue): Remove old comment.
2746 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2748 PR tree-optimization/30322
2749 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2750 (cand_value_at): Return the value as aff_tree.
2751 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2752 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2753 aff_combination_mult): New functions.
2754 (aff_combination_add): Use aff_combination_add_cst.
2755 (aff_combination_convert): Allow conversions to a wider type.
2756 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2757 * tree-affine.h (aff_combination_mult): Declare.
2759 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2761 * doc/tm.texi: Update documentation to reflect reality of exec
2762 and start file search behaviours. Update copyright year.
2763 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2764 header file directories.
2766 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2768 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2769 (*cosxf2): Rename to cosxf2_i387.
2770 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2771 using X87MODEF12 mode macro. Rename patterns to
2772 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2774 (*cosdf2, *cossf2): Ditto.
2775 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2776 to match extended input operands.
2777 (sincos<mode>3): New expander.
2778 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2779 insn patterns and corresponding splitters.
2781 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2783 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2784 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2786 2007-01-09 Nicolas Pitre <nico@cam.org>
2789 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2791 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2793 * target.h (struct gcc_target): New field library_rtti_comdat.
2794 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2795 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2796 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2797 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2799 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2801 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2803 2007-01-08 Richard Guenther <rguenther@suse.de>
2805 * tree.h (force_fit_type_double): Export.
2806 (force_fit_type): Remove.
2807 * fold-const.c (force_fit_type_double): New function.
2808 (force_fit_type): Remove.
2809 (int_const_binop): Use it.
2810 (fold_convert_const_int_from_int): Likewise.
2811 (fold_convert_const_int_from_real): Likewise.
2812 (fold_div_compare): Likewise.
2813 (fold_sign_changed_comparison): Likewise.
2814 (fold_unary): Likewise.
2815 (fold_negate_const): Likewise.
2816 (fold_abs_const): Likewise.
2817 (fold_not_const): Likewise.
2818 * c-common.c (shorten_compare): Use force_fit_type_double.
2819 * convert.c (convert_to_pointer): Likewise.
2821 2007-01-08 Richard Guenther <rguenther@suse.de>
2823 * tree.h (build_int_cst_wide_type): Export.
2824 * tree.c (build_int_cst_wide_type): New function.
2825 (build_int_cst_wide): Fix comment.
2826 * builtins.c (fold_builtin_object_size): Use build_int_cst
2827 to build -1 or 0 of the correct type. Use fit_double_type
2828 to check for overflow.
2829 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2831 (decode_field_reference): Likewise.
2832 (all_ones_mask_p): Likewise.
2833 (native_interpret_int): Use build_int_cst_wide_type.
2834 (fold_binary): Use build_int_cst_type to build an all-ones
2836 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2838 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2840 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2842 2007-01-09 Ben Elliston <bje@au.ibm.com>
2844 * genautomata.c (STATS_OPTION): New option.
2845 (stats_flag): New flag.
2846 (gen_automata_option): Handle it.
2847 (initiate_automaton_gen): Ditto.
2848 (write_automata): Output statistics only if stats_flag is
2849 set. Likewise, output time statistics only if time_flag is set.
2850 * doc/md.texi (Processor pipeline description): Document new flag.
2852 2007-01-08 Richard Guenther <rguenther@suse.de>
2854 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2855 * tree.c (build_int_cst_type): Likewise.
2856 (size_in_bytes): Don't call force_fit_type on the result.
2857 (int_fits_type_p): Use fit_double_type.
2858 * fold-const.c (fit_double_type): New function.
2859 (force_fit_type): Use it.
2860 * tree.h (fit_double_type): Export.
2862 2007-01-08 Jan Hubicka <jh@suse.cz>
2864 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2865 * ipa.c (function_and_variable_visibility): Fix return type.
2867 2007-01-08 Richard Guenther <rguenther@suse.de>
2869 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2870 of offset to build the index.
2871 * tree-pretty-print.c (dump_generic_node): Don't build negated
2872 const just for printing.
2873 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2874 * builtins.c (fold_builtin_int_roundingfn): Check if result
2875 fits the type by using force_fit_type and comparing the result.
2876 * predict.c (predict_loops): Use compare_tree_int for comparison.
2877 * tree.c (build_int_cst): Fall back to integer_type_node for
2879 (build_int_cst_wide): Assert type is non-null.
2881 2007-01-08 Roberto Costa <roberto.costa@st.com>
2883 * tree-vrp.c (extract_range_from_cond_expr): New.
2884 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2885 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2887 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2888 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2889 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2890 used as expressions.
2891 * tree-object-size.c (cond_expr_object_size): New.
2892 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2894 2007-01-08 Jan Hubicka <jh@suse.cz>
2896 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2897 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2900 2007-01-08 Richard Guenther <rguenther@suse.de>
2902 PR tree-optimization/23603
2903 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2904 (extract_range_from_binary): Fall back to truthvalue instead of
2905 varying for TRUTH_*_EXPR.
2906 (extract_range_from_comparison): Fall back to truthvalue instead of
2908 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2909 if all visited PHI values were constant.
2911 2007-01-08 Jan Hubicka <jh@suse.cz>
2913 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2914 (cgraph_analyze_function): break out from ...
2915 (cgraph_finalize_compilation_unit): ... here.
2916 (cgraph_expand_function): Remove forgoten commented out line.
2917 (cgraph_optimize): Analyze functions.
2919 2007-01-08 Jan Hubicka <jh@suse.cz>
2921 * tree-pas.h (TODO_remove_function): New flag.
2922 (TODO_update*): Renumber.
2923 (pass_ipa_increase_alignment,
2924 pass_ipa_function_and_variable_visibility): New passes.
2925 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2926 (cgraph_function_and_variable_visibility): Move to ipa.c
2927 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2928 cgraph_increase_alignment.
2929 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2930 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2931 calling tree-inline.
2932 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2933 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2934 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2935 (gate_increase_alignment): New function.
2936 (pass_ipa_increase_alignment): New pass.
2937 * ipa.c: Inline tree-pass.h and timevar.h
2938 (function_and_variable_visibility): Move here from cgraphunit.c
2939 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2940 * passes.c (init_optimization_passes): Add the two new passes.
2941 (execute_todo): Handle cgraph_remove_functions.
2943 2007-01-08 Nick Clifton <nickc@redhat.com>
2945 * config/frv/predicates.md (reg_or_0_operand): Accept
2948 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2950 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2951 * config.gcc: Add bfin*-rtems*.
2953 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2955 * c.opt: Add -flax-vector-conversions.
2956 * c-typeck.c (convert_for_assignment): Pass flag to
2957 vector_types_convertible_p to allow emission of note.
2958 (digest_init): Likewise.
2959 * c-opts.c: Handle -flax-vector-conversions.
2960 * c-common.c (flag_lax_vector_conversions): New.
2961 (vector_types_convertible_p): Unless -flax-vector conversions
2962 has been passed, disallow conversions between vectors with
2963 differing numbers of subparts and/or element types. If such
2964 a conversion is disallowed, possibly emit a note on the first
2965 occasion only to inform the user of -flax-vector-conversions.
2966 The new last argument specifies this.
2967 * c-common.h (flag_lax_vector_conversions): New.
2968 (vector_types_convertible_p): Add extra argument.
2969 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2970 char_type_node for V*QI type vectors.
2971 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2972 Update to satisfy new typechecking rules.
2973 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2975 * doc/invoke.texi (C Dialect Options): Document
2976 -flax-vector-conversions.
2978 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2980 PR tree-optimization/29877
2981 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2982 a register variable on the RHS to not be replaceable.
2984 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2985 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2986 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2987 const_simm15, const_pow2, const_npow2): Added.
2988 * config/score/misc.md (insv, extv, extzv, movmemsi,
2989 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2990 * config/score/score.c (score_address_cost, score_select_cc_mode):
2992 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2993 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2994 Update score_print_operand.
2995 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2996 Adjust register allocate order and update some macro define.
2997 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2998 mdx_block_move_straight, mdx_block_move_loop_head,
2999 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
3000 mdx_block_move): Added.
3001 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
3002 mdp_limm): Updated and fix some bug and typo.
3003 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
3004 (movsf, movdf, doloop_end): Added.
3006 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
3008 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
3009 config/arm/thumb2.md: Fix comment typos.
3010 * doc/extend.texi: Fix a typo.
3012 2007-01-07 Eric Christopher <echristo@apple.com>
3014 * configure.ac: Check for __stack_chk_fail for darwin.
3015 * configure: Regenerate.
3017 2007-01-07 Richard Guenther <rguenther@suse.de>
3019 * tree-vrp.c (extract_range_from_assert): CSE calls to
3020 compare_values where possible.
3021 (extract_range_from_unary_expr): Likewise.
3023 2007-01-07 Anatoly Sokolov <aesok@post.ru>
3025 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
3026 * config/avr/avr.c (call_insn_operand): Delete function.
3027 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
3028 instead of register number. Use predicates.
3029 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
3030 call_insn_operand): Add.
3032 2007-01-06 Jan Hubicka <jh@suse.cz>
3034 * tree-pass.h (pass_build_cgraph_edges): Declare.
3035 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
3036 (visited_nodes): Remove.
3037 (cgraph_create_edges): Move to cgraphbuild.c; rename to
3038 build_cgrpah_edges; make visited_nodes local.
3039 (cgraph_process_new_functions): DO not call initialize_inline_failed.
3040 (record_references_in_initializer): Move to cgraphbuild.c
3041 (initialize_inline_failed, rebuild_cgraph_edges,
3042 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
3043 (verify_cgraph_node): Make visited_nodes local.
3044 (cgraph_analyze_function): Do not call cgraph_create_edges and
3045 initialize_inline_failed.
3046 (cgraph_expand_function): Do not call cgraph_lower_function;
3047 assert that function is already lowered.
3048 * Makefile.in (cgraphbuild.o): New.
3049 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
3050 at the end of lowering passes.
3052 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
3054 * ifcvt.c (cond_move_convert_if_block): New function, code
3055 factored out from...
3056 (cond_move_process_if_block): ...here. Call the new function
3057 on the THEN and ELSE blocks.
3058 (merge_if_block): Do not copy global_live_at_end, merge_blocks
3059 already takes care of this.
3061 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3064 * tree.h (TREE_OVERFLOW_P): New.
3065 * c-typeck.c (parser_build_unary_op): Warn only if result
3066 overflowed and operands did not.
3067 (parser_build_binary_op): Likewise.
3068 (convert_for_assignment): Remove redundant overflow_warning.
3069 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
3071 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3073 * c-typeck.c (store_init_value): Split over two lines to follow
3074 the GNU coding style.
3076 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
3078 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
3079 __GXX_EXPERIMENTAL_CXX0X__.
3080 * doc/cpp.texi: Same.
3082 2007-01-05 Richard Guenther <rguenther@suse.de>
3085 * tree.c (get_narrower): Do not construct COMPONENT_REFs
3086 with mismatched types. Instead explicitly build a
3087 conversion NOP_EXPR.
3089 2007-01-05 Ian Lance Taylor <iant@google.com>
3091 * c-common.c (decl_with_nonnull_addr_p): New function.
3092 (c_common_truthvalue_conversion): Call it.
3093 * c-typeck.c (build_binary_op): Likewise.
3094 * c-common.h (decl_with_nonnull_addr_p): Declare.
3096 2007-01-05 Jakub Jelinek <jakub@redhat.com>
3099 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
3102 2007-01-05 Joel Brobecker <brobecker@adacore.com>
3104 * doc/install.texi (Final install): Document the fact that
3105 the GNAT runtime should not be stripped.
3107 2007-01-04 Jan Hubicka <jh@suse.cz>
3109 * tree-inline.c (fold_marked_statements): Update operand caches
3110 and EH after folding
3112 2007-01-04 Ian Lance Taylor <iant@google.com>
3114 * c-common.c (check_function_nonnull): Whitespace fix.
3116 2007-01-04 Jan Hubicka <jh@suse.cz>
3118 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
3119 comitted older version of patch.
3120 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
3121 function, celanup cfg and collect garbage.
3123 2007-01-04 Mike Stump <mrs@apple.com>
3125 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
3128 2007-01-04 Eric Christopher <echristo@apple.com>
3130 * libgcc2.c (__bswapsi2): Use SItype.
3131 (__bswapdi2): Use DItype.
3132 * libgcc2.h: Update for above.
3134 2007-01-04 Paul Brook <paul@codesourcery.com>
3136 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
3137 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
3138 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
3139 umulsidi3, umulsidi3adddi, smulsi3_highpart,
3140 umulsi3_highpart): Make conditional on !arm_arch6.
3141 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
3142 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
3143 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
3144 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
3145 umulsi3_highpart_v6): New insns.
3147 2007-01-04 Roger Sayle <roger@eyesopen.com>
3149 * fold-const.c (fold_convert): When casting an expression to void,
3150 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
3151 have a type. Instead of attempting to build a NOP_EXPR, return
3152 these "special" trees directly.
3154 2007-01-04 Joseph Myers <joseph@codesourcery.com>
3156 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
3157 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
3158 rs6000_cost->fp not 0 as adjustment for outer NEG.
3160 2007-01-04 Jan Hubicka <jh@suse.cz>
3162 * cgraph.c (cgraph_release_function_body): New function.
3163 (cgraph_remove_node): Use it.
3164 * cgraph.h (cgraph_release_function_body): Declare.
3165 * cgraphunit.c (cgraph_expand_function): Use it.
3166 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
3167 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
3168 is initialized and while compilation of other function is running.
3169 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
3170 statement CFG annotations from here to ...
3171 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
3173 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
3175 * cfgloop.h (enum li_flags): Make the constants powers of two.
3177 2007-01-04 Jan Hubicka <jh@suse.cz>
3179 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
3181 (fold_marked_statements): New function.
3182 (optimize_inline_calls, tree_function_versioning): Fold new statements.
3183 * tree-inline.h (copy_body_data): Add statements_to_fold.
3185 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
3187 * config.gcc: Mention libgcc/config.host.
3188 * Makefile.in: Update comments mentioning libgcc.
3189 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
3190 (all.cross, start.encap, rest.encap, rest.cross): Update
3191 dependencies for libgcc move.
3192 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
3193 (clean-target, clean-target-libgcc): Delete.
3194 (srcdirify, GCC_EXTRA_PARTS): New macros.
3195 (libgcc-support, libgcc.mvars): New rules.
3196 (distclean): Remove mention of mklibgcc.
3197 (install): Don't reference INSTALL_LIBGCC.
3198 (install-common): Don't reference EXTRA_PARTS.
3199 (install-libgcc, install-multilib): Delete rules.
3200 * mklibgcc.in: Delete file.
3201 * doc/configfiles.texi: Don't mention mklibgcc.
3203 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
3204 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
3205 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
3206 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
3207 (mklibgcc, ldblspecs): Likewise.
3209 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
3210 (posixpre.def): Use $(T).
3211 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
3212 (SHLIB_SRC, SHLIB_INSTALL): Delete.
3213 (SHLIB_LINK): Make dummy.
3214 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
3216 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
3218 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
3219 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
3220 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
3221 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
3223 * configure.ac (all_outputs): Remove mklibgcc.
3224 * configure: Regenerated.
3226 2007-01-03 Josh Conner <jconner@apple.com>
3229 * calls.c (compute_argument_addresses): Set stack and stack_slot
3230 for partial args, too.
3231 (store_one_arg): Use locate.size.constant for the size when
3232 generating a save_area.
3234 2007-01-03 Robert Kennedy <jimbob@google.com>
3236 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
3237 whose definitions are deleted due to basic block merging.
3239 2007-01-03 Paul Brook <paul@codesourcery.com>
3242 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
3244 (use_return_insn): Return 0 for Thumb interrupt functions.
3245 (print_multi_reg): Add rfe argument for IRQ returns.
3246 (arm_output_epilogue): Pop interrupt return address directly into PC.
3247 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
3249 2007-01-03 Paul Brook <paul@codesourcery.com>
3251 Merge from sourcerygxx-4_1.
3252 * config/arm/thumb2.md: New file.
3253 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
3255 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
3256 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3257 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
3258 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
3260 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3261 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
3263 * config/arm/ieee754-sf.S: Ditto.
3264 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
3265 (thumb1_base_register_rtx_p): ... to this.
3266 (thumb_index_register_rtx_p): Rename...
3267 (thumb1_index_register_rtx_p): ... to this.
3268 (thumb_output_function_prologue): Rename...
3269 (thumb1_output_function_prologue): ... to this.
3270 (thumb_legitimate_address_p): Rename...
3271 (thumb1_legitimate_address_p): ... to this.
3272 (thumb_rtx_costs): Rename...
3273 (thumb1_rtx_costs): ... to this.
3274 (thumb_compute_save_reg_mask): Rename...
3275 (thumb1_compute_save_reg_mask): ... to this.
3276 (thumb_final_prescan_insn): Rename...
3277 (thumb1_final_prescan_insn): ... to this.
3278 (thumb_expand_epilogue): Rename...
3279 (thumb1_expand_epilogue): ... to this.
3280 (arm_unwind_emit_stm): Rename...
3281 (arm_unwind_emit_sequence): ... to this.
3282 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3283 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3284 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3285 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3286 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3287 thumb2_output_casesi): New functions.
3288 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3289 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3290 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3291 THUMB2_WORK_REGS): Define.
3292 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3293 arm_condexec_mask, arm_condexec_masklen)): New variables.
3294 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3295 (arm_override_options): Check new CPU capabilities.
3296 Set new architecture flag variables.
3297 (arm_isr_value): Handle v7m interrupt functions.
3298 (user_return_insn): Return 0 for v7m interrupt functions. Handle
3300 (const_ok_for_arm): Handle Thumb-2 constants.
3301 (arm_gen_constant): Ditto. Use movw when available.
3302 (arm_function_ok_for_sibcall): Return false for v7m interrupt
3304 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3305 (thumb_find_work_register, arm_load_pic_register,
3306 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3307 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3308 print_multi_reg, output_mov_long_double_fpa_from_arm,
3309 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3310 output_mov_double_fpa_from_arm, output_move_double,
3311 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3312 output_return_instruction, arm_output_function_prologue,
3313 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3314 arm_output_mi_thunk, thumb_set_return_address): Ditto.
3315 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
3316 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3317 (arithmetic_instr, shift_op): Use arm_shift_nmem.
3318 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
3320 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3321 (thumb_expand_prologue): Use thumb_set_frame_pointer.
3322 (arm_file_start): Output directive for unified syntax.
3323 (arm_unwind_emit_set): Handle stack alignment instruction.
3324 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3325 Add v6t2, v7, v7a, v7r and v7m.
3326 (RETLDM): Add Thumb-2 code.
3327 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3328 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3329 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3330 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3331 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3332 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3333 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3334 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3335 ADJUST_INSN_LENGTH): Define.
3336 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3337 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3338 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3339 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3340 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3341 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3342 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3343 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3344 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3345 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3346 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3347 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3348 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3349 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3350 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3351 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3352 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3354 * config/arm/arm-tune.md: Regenerate.
3355 * config/arm/arm-protos.h: Update prototypes.
3356 * config/arm/vfp.md: Enable patterns for Thumb-2.
3357 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
3358 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3359 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3360 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3361 * config/arm/libunwind.S: Add Thumb-2 code.
3362 * config/arm/constraints.md: Update include Thumb-2.
3363 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3364 * config/arm/ieee754-df.S: Ditto.
3365 * config/arm/bpabi.S: Ditto.
3366 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3367 * config/arm/predicates.md (low_register_operand,
3368 low_reg_or_int_operand, thumb_16bit_operator): New.
3369 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3370 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3371 * config/arm/t-arm-elf: Add armv7 multilib.
3372 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3374 (UNSPEC_STACK_ALIGN, ce_count): New.
3375 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3376 arm_zero_extendsidi2, arm_zero_extendqidi2): New
3378 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3379 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3380 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3381 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3382 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3383 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3385 * doc/extend.texi: Document ARMv7-M interrupt functions.
3386 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3388 2007-01-03 Jakub Jelinek <jakub@redhat.com>
3390 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3391 (struct _Unwind_Context): Rename args_size to flags, remove
3392 signal_frame field, add a new args_size field and version field.
3393 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3394 _Unwind_IsExtendedContext): New inline functions.
3395 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3396 Assume by_value array is only present if _Unwind_IsExtendedContext.
3397 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3398 _Unwind_IsSignalFrame.
3399 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3400 (uw_update_context_1): Use _Unwind_SetSignalFrame.
3401 (uw_init_context_1): Initialize context->flags to
3402 EXTENDED_CONTEXT_BIT.
3403 * config/rs6000/linux-unwind.h (frob_update_context): Use
3404 _Unwind_SetSignalFrame.
3406 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
3409 * gimplify.c (gimplify_modify_expr_complex_part): Move below
3410 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
3413 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
3415 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3416 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3417 RET, AT_SP): Remove.
3418 * config/i386/i386.md (*sse_prologue_save_insn): Use return
3421 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3422 * cfgbuild.c (find_basic_blocks): Likewise.
3423 * cfgrtl.c (rtl_create_basic_block): Likewise.
3424 * function.c (temp_slots_at_level): Likewise.
3425 * reg-stack.c (stack_regs_mentioned): Likewise.
3426 * regclass.c (allocate_reg_info): Likewise.
3427 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3428 set_bb_for_stmt, move_block_to_fn): Likewise.
3429 * tree-complex.c (tree_lower_complex): Likewise.
3430 * vec.h (VEC_safe_grow_cleared): New.
3432 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3434 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
3436 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3437 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3439 * cfgloopmanip.c (scale_loop_frequencies): Export.
3440 (loopify): Scale the frequencies by prescribed coefficients.
3441 (set_zero_probability): New function.
3442 (duplicate_loop_to_header_edge): Improve updating of frequencies.
3443 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3444 and frequencies according to arguments.
3445 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3447 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3448 * modulo-sched.c (sms_schedule): Set probabilities for entering
3449 versioned loop correctly.
3450 * tree-vect-transform.c (vect_transform_loop): Ditto.
3451 * cfgloop.h (loopify, loop_version): Declaration changed.
3452 (scale_loop_frequencies): Declared.
3454 2007-01-02 Jan Hubicka <jh@suse.cz>
3456 * cgraph.c: Include tree-flow.h
3457 (cgraph_add_new-function): Handle IPA_SSA mode; execute
3459 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3460 * tree-pass.h (pass_all_early_optimizations): Declare.
3461 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3463 (cgraph_analyze_function): Do early_local_passes.
3464 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3465 Do not add referenced vars.
3466 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3468 (gate_all_early_local_passes): New gate.
3469 (pass_early_local_passes): Use new gate.
3470 (execute_early_local_optimizations): New functions.
3471 (gate_all_early_optimizations): New gate.
3472 (pass_all_early_optimizations): New pass.
3473 (execute_free_datastructures): Free SSA only when initialized.
3474 (gate_init_datastructures): Init only when optimizing.
3475 (tree_lowering_passes): Do early local passes when called late.
3476 * tree-profile.c (do_tree_profiling): Don't profile functions added
3478 (do_early_tree_profiling, pass_early_tree_profile): Kill.
3479 * tree-cfg.c (update_modified_stmts): Do not update when operands are
3481 * passes.c (init_optimizations_passes): Reorder so we go into SSA
3482 during early_local_passes.
3483 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3486 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3488 * Makefile.in: Update copyright year.
3490 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3492 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3493 in $(lang_checks) and check-consistency targets.
3495 2007-01-02 Jan Hubicka <jh@suse.cz>
3497 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3498 Do not add referenced vars.
3499 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3501 * passes.c (init_optimization_passes): Put mudflap_2 after
3502 free_datastructures.
3504 2007-01-02 Jan Hubicka <jh@suse.cz>
3506 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3507 Set NOTHROW flag on call statements proved to be nothrow.
3508 Update statement of local calls so new pure/const functions are
3509 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3511 (tree_rest_of_compilation): Register hooks and initialize bitmap
3512 early. Do not set after_inlining flag.
3514 2007-01-02 Steve Ellcey <sje@cup.hp.com>
3516 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
3517 HOST_BITS_PER_LONGLONG
3519 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3522 * c-typeck.c (store_init_value): Don't emit pedantic overflow
3523 warning for non-static initializers.
3525 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
3527 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3528 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3529 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3530 Always use set_unique_reg_note to add REG_EQUAL notes.
3532 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3535 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3537 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3538 * cfgbuild.c (find_basic_blocks): Likewise.
3539 * cfgrtl.c (rtl_create_basic_block): Likewise.
3540 * function.c (temp_slots_at_level): Likewise.
3541 * reg-stack.c (stack_regs_mentioned): Likewise.
3542 * regclass.c (allocate_reg_info): Likewise.
3543 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3544 set_bb_for_stmt, move_block_to_fn): Likewise.
3545 * tree-complex.c (tree_lower_complex): Likewise.
3546 * vec.h (VEC_safe_grow_cleared): New.
3548 2007-01-02 Ian Lance Taylor <iant@google.com>
3550 * c-common.c (c_common_truthvalue_conversion): When warning about
3551 using an assignment as a truth value, set TREE_NO_WARNING.
3553 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3556 * c.opt (Wold-style-declaration): New.
3557 * doc/invoke.texi (C-only Warning Options): New.
3558 (Wold-style-declaration): Document it.
3559 (Wextra): Enabled by -Wextra.
3560 * c-opts.c (c_common_post_options): Enabled by -Wextra.
3561 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3562 -Wold-style-declaration.
3564 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3566 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3567 * cfgbuild.c (find_basic_blocks): Likewise.
3568 * cfgrtl.c (rtl_create_basic_block): Likewise.
3569 * function.c (temp_slots_at_level): Likewise.
3570 * reg-stack.c (stack_regs_mentioned): Likewise.
3571 * regclass.c (allocate_reg_info): Likewise.
3572 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3573 set_bb_for_stmt, move_block_to_fn): Likewise.
3574 * tree-complex.c (tree_lower_complex): Likewise.
3575 * vec.h (VEC_safe_grow_cleared): New.
3577 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
3579 * c-common.c (c_common_nodes_and_builtins): Since variants of
3580 void_type_node get built before it is given a name, we need to
3581 give those variants the name, too.
3582 (complete_array_type): We need to work with the canonical main
3583 type of the array, from which we will build the qualified version.
3584 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3585 * print-tree.c (print_node): Display canonical type information
3587 * stor-layout.c (layout_type): When we don't know the
3588 alignment of a type for which we're building an array, we end up
3589 guessing wrong, so make the type require structural equality.
3590 * tree.c (make_node_stat): When we build a new type, it is its
3592 (build_type_attribute_qual_variant): When building an attribute
3593 variant, its canonical type is the non-attribute variant. However,
3594 if the attributes are target-dependent and they differ, we need to
3595 use structural equality checks for this type.
3596 (build_qualified_type): A qualified type is not equivalent to its
3597 unqualified variant; set the canonical type appropriately.
3598 (build_distinct_type_copy): When building a distinct type from
3599 another type, the new type is its own canonical type.
3600 (build_variant_type_copy): When building a new type variant, we
3601 assume that it is equivalent to the original type.
3602 (build_pointer_type_for_mode): When building a pointer type, also
3603 build a canonical type pointer.
3604 (build_reference_type_for_mode): When building a reference type,
3605 also build a canonical type reference.
3606 (build_index_type): When we can't hash an index type (e.g.,
3607 because its maximum value is negative), the index type requires
3608 structural equality tests.
3609 (build_array_type): Build the canonical form of an array type.
3610 (build_function_type): Function types require structural equality,
3611 because they contain default arguments, attributes, etc.
3612 (build_method_type_directly): Ditto for method types.
3613 (build_offset_type): Build the canonical offset type.
3614 (build_complex_type): Build the canonical vector type.
3615 (make_vector_type): Build the canonical vector type.
3616 * tree.h (TYPE_CANONICAL): New.
3617 (TYPE_STRUCTURAL_EQUALITY_P): New.
3618 (SET_TYPE_STRUCTURAL_EQUALITY): New.
3619 (struct tree_type): Added "canonical" field.
3620 * params.h (VERIFY_CANONICAL_TYPES): New.
3621 * doc/c-tree.texi (TYPE_CANONICAL): Document.
3622 (TYPE_STRUCTURAL_EQUALITY_P): Document.
3623 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3624 * doc/invoke.texi (verify-canonical-types): Document --param
3625 parameter for verifying canonical types.
3627 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3629 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3630 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3631 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3632 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3634 * config/rs6000/e500.h: New.
3635 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3636 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3638 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3639 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3640 TARGET_E500_DOUBLE): Remove.
3641 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3642 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3643 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3644 * config/rs6000/rs6000.c (rs6000_override_options): Use
3647 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3649 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3650 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3651 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3652 rs6000_emit_cmove): Don't check TARGET_E500.
3653 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3654 sordered): Don't check TARGET_E500.
3656 2007-01-01 Eric Christopher <echristo@apple.com>
3658 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3659 HARD_REGISTER_NUM_P.
3661 2007-01-01 Roger Sayle <roger@eyesopen.com>
3663 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3664 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3665 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3668 2007-01-01 Mike Stump <mrs@apple.com>
3670 * configure.ac: Remove support for building with Apple's gcc-3.1.
3672 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3675 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3676 REG is a hard register. Do not modify REG before calling
3678 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3679 multiple of the size of YMODE for certain lowpart cases.
3681 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3684 * gimplify (voidify_wrapper_expr): Update for
3687 2007-01-01 Andreas Schwab <schwab@suse.de>
3690 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3691 save of BR0 in extra_spill_size instead of spill_size.
3692 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3694 (ia64_expand_epilogue): Restore BR0 from its new location.
3696 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3698 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3699 Use a temporary variable if the left hand side is not a gimple
3702 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3704 * gimplify.c (gimplify_return_expr): Make the temporary variable
3705 for the return expression, a gimple register variable.
3707 2007-01-01 Jan Hubicka <jh@suse.cz>
3709 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3712 2007-01-01 Mike Stump <mrs@apple.com>
3714 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3716 * config/darwin.c (machopic_select_rtx_section): Use
3718 (darwin_mergeable_constant_section): Likewise.
3719 * configure: Regenerate.
3720 * config.in: Regenerate.
3722 2007-01-01 Jan Hubicka <jh@suse.cz>
3723 Andrew Pinski <pinskia@gmail.com>
3725 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3726 before starting IPA passes.