gcc/ChangeLog

   1 2007-01-30  Dirk Mueller  <dmueller@suse.de>
   2
   3         PR c++/30601
   4         * doc/invoke.texi (-Wreturn-type): Update description to
   5         match new behavior.
   6
   7 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
   8
   9         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
  10
  11 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
  12
  13         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
  14         to handle optabs that operate on floating point input argument and
  15         output to integer output.
  16         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
  17         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
  18         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
  19         expand_builtin_interclass_mathfn ().
  20         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
  21         (ilogbsi2): Remove.
  22         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
  23         ilogbl built-in functions as x87 intrinsics.
  24
  25 2007-01-30  Richard Guenther  <rguenther@suse.de>
  26
  27         PR middle-end/27657
  28         * dwarf2out.c (reference_to_unused): Query varpool if the
  29         variable was output.
  30
  31 2007-01-30  Richard Guenther  <rguenther@suse.de>
  32
  33         PR middle-end/30313
  34         * passes.c (execute_one_pass): Reset in_gimple_form to not
  35         confuse non-unit-at-a-time mode.
  36
  37 2007-01-29  Roger Sayle  <roger@eyesopen.com>
  38             Richard Guenther  <rguenther@suse.de>
  39
  40         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
  41         avoid undefined behaviour on overflow.  Use force_fit_type_double
  42         to construct the constant with the specified TREE_OVERFLOW.
  43
  44 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
  45
  46         * config/dfp-bit.c: Add parameterized support for fp exceptions.
  47         * config/dfp-bit.h: Ditto.
  48
  49 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
  50
  51         * c-decl.c (pop_scope): Replace warnings with call to
  52         warn_for_unused_label.
  53         * c-common.h (warn_for_unused_label): Declare.
  54         * c-common.c (warn_for_unused_label): Define.
  55
  56 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
  57
  58         * tree-optimize.c (update_inlined_to_pointers): Delete unused
  59         function.
  60
  61 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
  62
  63         * Makefile.in (USER_H): Remove decfloat.h.
  64         * ginclude/decfloat.h: Delete, moving contents to ...
  65         * ginclude/float.h: Add support for decimal floating point,
  66         guarded by __STDC_WANT_DEC_FP__.
  67
  68 2007-01-29  Mike Stump  <mrs@apple.com>
  69
  70         * doc/gccint.texi (Top): Rename Loop Representation to Loop
  71         Analysis and Representation to resolve case insensitive conflict.
  72         * doc/loop.texi (Loop Analysis and Representation): Likewise.
  73
  74 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
  75
  76         * tree.h (struct tree_memory_tag): Add aliases member.
  77         (MTAG_ALIASES): New macro.
  78         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
  79         (add_may_alias): Remove pointer-set. Update for may_aliases being
  80         a bitmap.
  81         (mark_aliases_call_clobbered): Update for may_aliases being a
  82         bitmap.
  83         (compute_tag_properties): Ditto.
  84         (create_partition_for): Ditto.
  85         (compute_memory_partitions): Ditto.
  86         (dump_may_aliases_for): Ditto.
  87         (is_aliased_with): Ditto.
  88         (add_may_alias_for_new_tag): Ditto.
  89         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
  90         (compute_is_aliased): New function.
  91         (compute_may_aliases): Call compute_is_aliased).
  92         (init_alias_info): Initialize alias_bitmap_obstack.
  93         (union_alias_set_into): New function.
  94         (compute_flow_sensitive_aliasing): Use union_aliases_into.
  95         (have_common_aliases_p): Rewrite to take two bitmaps and use
  96         intersection.
  97         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
  98         Update for bitmaps.
  99         (finalize_ref_all_pointers): Update for add_may_alias changes.
 100         (new_type_alias): Ditto.
 101         * tree-flow-inline.h (may_aliases): Return a bitmap.
 102         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
 103         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
 104         may_aliases being a bitmap.
 105         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
 106         may_aliases now returns a bitmap.
 107         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
 108         being a bitmap.
 109         * tree-ssa-operands.c (add_virtual_operand): Update for
 110         may_aliases being a bitmap.
 111
 112 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
 113
 114         PR tree-optimization/30630
 115         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
 116         variable as changed.
 117
 118 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
 119
 120         PR c++/28266
 121         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
 122         expanded only once even if an error occurs.
 123
 124 2007-01-29  Ben Elliston  <bje@au.ibm.com>
 125
 126         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
 127         (__gcov_average_profiler): Likewise.
 128         (__gcov_ior_profiler): Likewise.
 129         (__gcov_merge_ior): Likewise.
 130
 131 2007-01-28  Jan Hubicka  <jh@suse.cz>
 132
 133         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
 134         * value-prof.c (stringop_block_profile): Likewise.
 135
 136 2007-01-28  Jan Hubicka  <jh@suse.cz>
 137
 138         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
 139         variant handling histograms; add wrapper.
 140         (clear_storage_via_libcall): Export.
 141         (emit_block_move_hints): Break out from ...; add histograms.
 142         (emit_block_move): ... this one.
 143         (clear_storage_hints): Break out from ...; add histograms.
 144         (clear_storage): ... this one.
 145         (set_storage_via_memset): Handle histogram.
 146         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
 147         (clear_storage_hints, clear_storage_via_libcall): Declare.
 148         (set_storage_via_setmem): Update prototype.
 149         * doc/md.texi (movmem, setmem): Document new arguments.
 150
 151         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
 152         new histograms.
 153         (stringop_block_profile): New global function.
 154         (tree_stringops_values_to_profile): Profile block size and alignment.
 155         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
 156         HIST_TYPE_IOR.
 157         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
 158         (stringop_block_profile): Declare.
 159         * builtins.c: Include value-prof.h.
 160         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
 161         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
 162         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
 163         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
 164         * profile.c (instrument_values): Add new counters.
 165         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
 166         calls.
 167         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
 168         (tree_init_edge_profiler): Build new profilers.
 169         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
 170         (pass_tree_profile): Add dump.
 171         (tree_profile_hooks): Update.
 172         * Makefile.in (LIBGCOV): Add new constants.
 173         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
 174         __gcov_ior_profiler): New.
 175         * i386.md (movmem/setmem expanders): Add new optional arguments.
 176
 177 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
 178
 179         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
 180
 181 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
 182
 183         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
 184         (tree_unswitch_single_loop): Pass eni_size_weights to
 185         tree_num_loop_insns.
 186         * tree-ssa-loop-manip.c: Include tree-inline.h.
 187         (can_unroll_loop_p): Pass eni_size_weights to
 188         tree_num_loop_insns.
 189         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
 190         Pass eni_size_weights to estimate_num_insns.
 191         * tree.h (init_inline_once): Export.
 192         * toplev.c (backend_init): Call init_inline_once.
 193         * cgraphunit.c (cgraph_process_new_functions,
 194         cgraph_analyze_function): Pass eni_inlining_weights to
 195         estimate_num_insns.
 196         * ipa-inline.c (compute_inline_parameters): Ditto.
 197         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
 198         to estimate_num_insns.
 199         (try_unroll_loop_completely): Pass eni_size_weights to
 200         tree_num_loop_insns.
 201         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
 202         ot estimate_num_insns.
 203         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
 204         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
 205         * tree-inline.c (eni_inlining_weights, eni_size_weights,
 206         eni_time_weights): New variables.
 207         (init_inline_once): Initialize them.
 208         (struct eni_data): Mew.
 209         (estimate_num_insns_1, estimate_num_insns): Use weights.
 210         * tree-inline.h (struct eni_weights_d): New.
 211         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
 212         (estimate_num_insns): Declaration changed.
 213         * cfgloop.h (tree_num_loop_insns): Declaration changed.
 214         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
 215         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
 216
 217 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
 218
 219         * tree-data-ref.c (conflict_fn): Assert that the number of affine
 220         relations in the conflict function is valid.
 221
 222 2007-01-27  Ian Lance Taylor  <iant@google.com>
 223
 224         * common.opt: Add fstrict-overflow.
 225         * opts.c (decode_options): Set flag_strict_overflow if -O2.
 226         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
 227         (TYPE_OVERFLOW_UNDEFINED): Define.
 228         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
 229         Replace all uses.
 230         * tree.h (TYPE_TRAP_SIGNED): Don't define.
 231         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
 232         (fold_negate_expr): Likewise.
 233         (make_range): Likewise.
 234         (extract_muldiv_1): Likewise.
 235         (maybe_canonicalize_comparison): Likewise.
 236         (fold_comparison): Likewise.
 237         (fold_binary): Likewise.
 238         (tree_expr_nonnegative_p): Likewise.
 239         (tree_expr_nonzero_p): Likewise.
 240         * tree-vrp.c (compare_values): Likewise.
 241         (extract_range_from_binary_expr): Likewise.
 242         (extract_range_from_unary_expr): Likewise.
 243         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
 244         Likewise.
 245         (nowrap_type_p): Likewise.
 246         * tree-scalar-evolution.c (simple_iv): Likewise.
 247         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
 248         (build_range_check): Likewise.
 249         (extract_muldiv_1): Likewise.
 250         (fold_comparison): Likewise.
 251         * tree-vrp.c (vrp_int_const_binop): Likewise.
 252         (extract_range_from_unary_expr): Likewise.
 253         * convert.c (convert_to_integer): Likewise.
 254         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
 255         (fold_comparison): Likewise.
 256         (fold_binary): Likewise.
 257         * optabs.c (optab_for_tree_code): Likewise.
 258         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
 259         * simplify-rtx.c (simplify_const_relational_operation): Check
 260         flag_strict_overflow and flag_trapv.
 261         (simplify_const_relational_operation): Likewise.
 262         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
 263         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
 264         -fstrict-overflow.
 265
 266 2007-01-27  Roger Sayle  <roger@eyesopen.com>
 267
 268         * tree.c (tree_fold_gcd): Delete.
 269         * tree.h (tree_fold_gcd): Remove prototype.
 270         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
 271         test whether one constant integer is a multiple of another.  Instead
 272         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
 273         * fold-const.c (multiple_of_p):  We've determined both TOP and
 274         BOTTOM are integer constants so we can call int_const_binop directly
 275         instead of the more generic const_binop.
 276
 277 2007-01-27  Roger Sayle  <roger@eyesopen.com>
 278
 279         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
 280         1*X check that the constant hasn't overflowed, to preserve the
 281         TREE_OVERFLOW bit.
 282         (round_up): Provide an efficient implementation when rouding-up an
 283         INTEGER_CST to a power-of-two.
 284
 285 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
 286
 287         * doc/sourcebuild.texi: Add comma for clarity.
 288         * doc/extend.texi: Fix some typos.
 289         * doc/passes.texi: Likewise.
 290         * doc/cppinternals.texi: Likewise.
 291         * doc/c-tree.texi: Likewise.
 292         * doc/tree-ssa.texi: Likewise.
 293         * doc/install.texi: Likewise.
 294
 295 2007-01-27  Jan Hubicka  <jh@suse.cz>
 296
 297         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
 298         (pass_sra): Do not require alias information.
 299         * passes.c (init_optimization_passes): Add SRA
 300
 301 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
 302
 303         * tracer.c (rest_of_handle_tracer): We already cleaned
 304         up the CFG in tracer() so don't do it here again.
 305         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
 306         cleanup_cfg here, either.  And don't call renumber_insns.
 307
 308         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
 309         and BB_HEAD are in the insn stream here.  Instead make sure
 310         that BB_INSN is valid on all insns.  Also, do check here that
 311         there are no pending branch predictions...
 312         (rtl_verify_flow_info): ...instead of doing it here.  Checks
 313         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
 314         here.
 315
 316 2007-01-26  Roger Sayle  <roger@eyesopen.com>
 317
 318         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
 319         function to simplify/factorize operand order canonicalization.
 320         (ix86_fixup_binary_operands): Reorganize using the above function.
 321         (ix86_binary_operator_ok): Likewise.
 322
 323 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
 324
 325         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
 326         file scope from optimize_attrs.
 327         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
 328         up search.
 329         (optimize_attrs): Clear insn_code_values after freeing it.
 330
 331 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
 332
 333         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
 334         a parts.base assignment.
 335
 336 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
 337
 338         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
 339         (compute_subscript_distance, initialize_data_dependence_relation,
 340         finalize_ddr_dependent, analyze_ziv_subscript,
 341         analyze_siv_subscript_cst_affine,
 342         compute_overlap_steps_for_affine_univar,
 343         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
 344         analyze_siv_subscript, analyze_miv_subscript,
 345         analyze_overlapping_iterations, subscript_dependence_tester_1,
 346         compute_self_dependence, free_dependence_relation): Work
 347         with affine_fn instead of chrecs.
 348         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
 349         common_affine_function, affine_function_base,
 350         affine_function_constant_p, affine_fn_op, affine_fn_plus,
 351         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
 352         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
 353         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
 354         (all_chrecs_equal_p): Removed.
 355         * tree-data-ref.h (affine_fn, conflict_function): New types.
 356         (struct subscript): Change type of conflicting_iterations_in_a
 357         and conflicting_iterations_in_b.
 358
 359 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
 360
 361         PR other/30182
 362         * config/pa/pa.h (TARGET_HPUX_11): New.
 363         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
 364         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
 365
 366 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
 367         Richard Guenther  <rguenther@suse.de>
 368
 369         * tree-ssa-structalias.c (solve_graph): Handle case
 370         we merged the variable to another.
 371
 372 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
 373
 374         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
 375         functions.
 376
 377         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
 378         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
 379
 380 2007-01-25  DJ Delorie  <dj@redhat.com>
 381
 382         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
 383         allow changes to modes which don't fit in those registers.
 384
 385         * reload1.c (choose_reload_regs): Check for invalid subregs before
 386         computing their locations, not after.
 387
 388 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
 389
 390         PR 25127
 391         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
 392         save Altivec registers in an eh_return function.
 393         (compute_vrsave_mask): Likewise.
 394         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
 395         for saving Altivec registers.
 396         (rs6000_emit_prologue): Don't allocate stack twice in
 397         eh_return function.  Correct expected value of altivec_save_offset
 398         when using save_world.  Describe save of R0 to stack when using
 399         save_world.  Describe stack pointer adjustment when using
 400         save_world.  Remove duplicated eh_return parameter register saving.
 401         Update sp_offset variable after save_world.
 402         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
 403         darwin-world.asm.
 404         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
 405         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
 406         implies Altivec.
 407
 408 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
 409
 410         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
 411
 412 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
 413
 414         PR other/30182
 415         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
 416         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
 417
 418 2007-01-25  Richard Guenther  <rguenther@suse.de>
 419
 420         * doc/invoke.texi (-Wcoverage-mismatch): Document.
 421         * common.opt (-Wcoverage-mismatch): New warning option.
 422         * coverage.c (get_coverage_counts): Ignore coverage mismatch
 423         if -Wcoverage-mismatch is given.
 424
 425 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
 426
 427         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
 428         (ipcp_driver): Change to static definition.
 429         Add dumping of the ifunctions.
 430         (constant_val_insert): Remove unused parameter. Support for SSA.
 431         (ipcp_propagate_const): Support for SSA.
 432         (ipcp_profile_bb_print): Print only analyzed nodes.
 433         (ipcp_replace_map_create): Remove support for Fortran constant
 434         for now.
 435         * ipa-prop.c (ipa_method_modify_stmt,
 436         ipa_callsite_compute_param): Support for SSA.
 437         * ipa-prop.h (ipcp_driver): Remove declaration.
 438         (IS_VALID_TREE_MAP_INDEX): Add define.
 439
 440 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
 441
 442         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
 443
 444 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
 445
 446         PR middle-end/30447
 447         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
 448         constant argument to cabs and do it without checking for
 449         -funsafe-math-optimizations.
 450
 451 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
 452
 453         * c-common.h (RID_FIRST_CXX0X): New.
 454         (RID_LAST_CXX0X): New.
 455         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
 456         by -Wall.
 457         * c.opt (Wc++0x-compat): New.
 458         * doc/invoke.texi (-Wc++0x-compat): Document.
 459
 460 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
 461
 462         * builtins.c (fold_builtin_carg): New.
 463         (fold_builtin_1): Use it.
 464
 465 2007-01-24  Jan Hubicka  <jh@suse.cz>
 466
 467         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
 468
 469         * ipa-inline.c (initial_insns, max_insns): Delete.
 470         (compute_max_insns): New function.
 471         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
 472         of insns as base for code growth.
 473         (cgraph_decide_inlining): Make initial_insns local; do not compute
 474         max_insns.
 475         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
 476         * doc/invoke.texi (inline-unit-growth): Update docs.
 477
 478 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
 479
 480         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
 481         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
 482         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
 483         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
 484         for CPUs that have PTA_CX16 set.
 485
 486         PR middle-end/27416
 487         * gimplify.c (omp_check_private): New function.
 488         (gimplify_scan_omp_clauses): Use it for
 489         firstprivate/lastprivate/reduction.
 490
 491         PR middle-end/30494
 492         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
 493         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
 494
 495         PR middle-end/30421
 496         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
 497         If lower_lastprivate_clauses emits some statements, append them
 498         to dlist rather than body_p and to body_p append an initializer.
 499         (lower_omp_for): Adjust caller.
 500
 501 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
 502
 503         * target.h (globalize_decl_name): New.
 504         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
 505         * output.h (default_globalize_decl_name): New.
 506         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
 507         globalize_label.
 508         (globalize_decl): Ditto.
 509         (default_globalize_decl_name): New.
 510         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
 511         (ia64_handle_version_id_attribute): New.
 512         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
 513         (ia64_asm_output_external): Use globalize_decl_name instead
 514         of globalize_label.
 515         * doc/extend.texi (version_id): New pragma.
 516         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
 517
 518 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
 519
 520         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
 521         _uleb128_t and _Unwind_SWord with _sleb128_t.
 522         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
 523         uw_frame_state_for, uw_update_context_1): Likewise.
 524         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
 525         * unwind-pe.h (read_uleb128, read_sleb128,
 526         read_encoded_value_with_base): Likewise.
 527         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
 528
 529 2007-01-24  Richard Guenther  <rguenther@suse.de>
 530
 531         * builtins.c (expand_builtin_cexpi): Get the fndecl
 532         for cexp in the correct way.
 533
 534 2007-01-24  Jan Hubicka  <jh@suse.cz>
 535
 536         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
 537
 538 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
 539
 540         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
 541         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
 542         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
 543         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
 544         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
 545         __SIZEOF_PTRDIFF_T__.
 546         * doc/cpp.texi: Documentation for the new builtin macros added.
 547
 548 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
 549
 550         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
 551         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
 552         instead of operands[2] to avoid access past the end of array.
 553
 554 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
 555
 556         * reload1.c (emit_reload_insns): Pass the reload register
 557         for a non-spill output reload through forget_old_reloads_1.
 558
 559 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
 560
 561         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
 562         _SOFT_DOUBLE if doubles use software floating-point.
 563         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
 564         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
 565         * config/rs6000/darwin-ldouble.c: Also compile functions for
 566         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
 567         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
 568         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
 569         commented-out long double override.
 570         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
 571         * config/rs6000/eabispe.h: Likewise.
 572         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
 573         long double for non-SPE.
 574         (rs6000_handle_option): Likewise.
 575         (invalid_e500_subreg): Disallow more subregs involding DImode,
 576         DFmode, TImode or TFmode.
 577         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
 578         E500 double.
 579         (legitimate_lo_sum_address_p): Also check for TFmode for E500
 580         double.
 581         (rs6000_legitimize_address): Also handle TFmode for E500 double.
 582         (rs6000_legitimize_reload_address): Also handle TFmode for E500
 583         double.
 584         (rs6000_legitimate_address): Also check for TFmode for E500
 585         double.
 586         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
 587         (spe_build_register_parallel): Handle TFmode and TCmode.
 588         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
 589         double.
 590         (function_arg): Handle TFmode and TCmode for E500 double.
 591         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
 592         in general.
 593         (print_operand): Handle TFmode and TImode for %y.
 594         (rs6000_generate_compare): Handle TFmode comparisons for E500
 595         double.
 596         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
 597         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
 598         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
 599         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
 600         TFmode for E500 double.
 601         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
 602         (floatsidf2): Enable for E500 double.
 603         (movtf_softfloat): Use rs6000_nonimmediate_operand.
 604         (extenddftf2): Change to extenddftf2_fprs.
 605         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
 606         depending on TARGET_E500_DOUBLE.
 607         (extendsftf2): Enable for E500 double.
 608         (trunctfdf2): Enable for E500 double.
 609         (trunctfsf2): Change to trunctfsf2_fprs.
 610         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
 611         depending on TARGET_E500_DOUBLE.
 612         (floatsitf2): Enable for E500 double.
 613         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
 614         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
 615         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
 616         (negtf2): Change to negtf2_internal.
 617         (negtf2): New expander.
 618         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
 619         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
 620         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
 621         (movdi_internal32): Use rs6000_nonimmediate_operand.
 622         (unnamed splitter): Likewise.
 623         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
 624         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
 625         (SPE64TF, DITI): New mode macros.
 626         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
 627         modes.
 628         (frob_tf_ti): New.
 629         (frob_<mode>_di_2): New.
 630         (frob_tf_di_8_2): New.
 631         (frob_di_df): Change to frob_di_<mode>; allow more modes.
 632         (frob_ti_tf): New.
 633         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
 634         more modes.
 635         (frob_ti_<mode>_8_2): New.
 636         (frob_ti_tf_2): New.
 637         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
 638         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
 639         TFmode.
 640         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
 641         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
 642         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
 643         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
 644         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
 645         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
 646         tsttflt_gp): New.
 647
 648 2007-01-23  Ian Lance Taylor  <iant@google.com>
 649
 650         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
 651         insn-*.o first.
 652         (OBJS-archive): Reformat, alphabetize.
 653         (OBJS): Change out_object_file to OBJS-md.
 654
 655 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
 656
 657         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
 658         JUMP_P and CALL_P predicates where applicable.
 659         * config/i386/i386.c: Ditto.
 660         * config/i386/i386.md: Ditto.
 661         * config/i386/mmx.md: Ditto.
 662         * config/i386/predicates.md: Ditto.
 663
 664 2007-01-22  Andreas Schwab  <schwab@suse.de>
 665
 666         * config/m68k/m68k.h: Fix comment.
 667
 668 2007-01-22  Jan Hubicka  <jh@suse.cz>
 669
 670         * passes.c (init_optimization_passes): Do not rerun
 671         pass_early_warn_uninitialized.
 672
 673 2007-01-22  Richard Guenther  <rguenther@suse.de>
 674
 675         PR tree-optimization/30038
 676         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
 677         function.
 678         (execute_cse_sincos_1): Likewise.
 679         (execute_cse_sincos): Likewise.
 680         (gate_cse_sincos): Likewise.
 681         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
 682         the cexpi() canonicalization of sincos().
 683         * tree-pass.h (pass_cse_sincos): Declare.
 684         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
 685
 686 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
 687
 688         PR rtl-optimization/29329
 689         * combine.c (replaced_rhs_insn): Rename to i2mod.
 690         (replaced_rhs_value): Rename to i2mod_new_rhs.
 691         (i2mod_old_rhs): New global variable.
 692         (combine_instructions): Adjust for above change.  Save a copy of
 693         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
 694         note are substituted in the second instruction.
 695         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
 696         ditch the note if it pertains to the second eliminated register
 697         and this register is mentioned in i2mod_old_rhs.
 698
 699         Revert:
 700         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
 701
 702         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
 703         the insn to which the note was originally attached.
 704
 705 2007-01-21  Jan Hubicka  <jh@suse.cz>
 706
 707         * ipa-inline.c (inlining_mode): Comment, move up.
 708         (cgraph_decide_inlining_incrementally): Do not perform inlining
 709         itself; fix handling of flattening of self recursive functions.
 710         (cgraph_find_cycles): Remove.
 711         (cgraph_flatten_node): Remove.
 712         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
 713         (try_inline): New function.
 714         (cgraph_early_inlining): Update call of
 715         cgraph_decide_inlining_incrementally.  Apply inlining here.
 716         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
 717
 718 2007-01-21  Dirk Mueller  <dmueller@suse.de>
 719
 720         PR bootstrap/30511
 721         * tree-vrp.c (check_array_bounds): do not warn
 722         about ADDR_EXPR's of ARRAY_REF's which are immediately
 723         used in binary expressions.
 724
 725 2007-01-21  Ira Rosen  <irar@il.ibm.com>
 726
 727         * tree-vectorizer.h (struct _stmt_vec_info): Add new field
 728         read_write_dep and macros for its access.
 729         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
 730         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
 731         argument, call vect_check_interleaving for every independent pair of
 732         data-refs. Mark loads that access the same memory location as a store
 733         in the loop.
 734         (vect_check_dependences): Remove.
 735         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences
 736         call, fix the call to vect_analyze_data_ref_dependence.
 737         (vect_analyze_data_ref_access): For statements that access the same
 738         data-ref, check that they are not stores; for loads, check that there
 739         is no store that access the same location.
 740
 741 2007-01-20  Roger Sayle  <roger@eyesopen.com>
 742             Joseph Myers  <joseph@codesourcery.com>
 743
 744         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
 745         (-fdump-rtl-stack): Likewise.
 746         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
 747         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
 748         (-mpt-fixed): Use @minus{} for minus sign.
 749         (Using Precompiled Headers): Fix under/overfull hbox.
 750
 751 2007-01-20  Jan Hubicka  <jh@suse.cz>
 752
 753         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
 754         place within annotation.
 755
 756 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
 757
 758         * pa.c (output_move_double): Change array size of xoperands to 4.
 759
 760 2007-01-20  Jan Hubicka  <jh@suse.cz>
 761
 762         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
 763         as temporary.
 764
 765 2007-01-19  Ian Lance Taylor  <iant@google.com>
 766
 767         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
 768         constant to unsigned HOST_WIDE_INT before negating.
 769
 770 2007-01-19  Ian Lance Taylor  <iant@google.com>
 771
 772         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
 773         to unsigned.
 774         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
 775         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
 776         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
 777         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
 778         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
 779         (realloc_vdef, realloc_vuse): Change second parameter to
 780         unsigned.
 781         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
 782         fields to unsigned.
 783         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
 784         unsigned. Change x and lim locals to unsigned.
 785         (realloc_vdef, realloc_vuse): Change num_elem parameter to
 786         unsigned.
 787         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
 788         (copy_virtual_operands): Change i and n locals to unsigned.
 789
 790 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
 791
 792         PR middle-end/29335
 793         * builtins.c (fold_builtin_1): Handle builtin fdim.
 794
 795 2007-01-20  Jan Hubicka  <jh@suse.cz>
 796
 797         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
 798         * tree-ssa-structalias.c (compute_points_to_sets): Do call
 799         init_alias_heapvars.
 800         (init_alias_heapvars): Initialize only when not already initialized.
 801         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
 802
 803 2007-01-19  Roger Sayle  <roger@eyesopen.com>
 804
 805         * common.opt (fsigned-zeros): New command line option.
 806         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
 807         of flag_unsafe_math_optimizations.
 808         * opts.c (set_fast_math_flags): The -ffast-math command line option
 809         implies -fno-signed-zeros.
 810         (fast_math_flags_set_p): Likewise.
 811
 812         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
 813         the documentation of -ffast-math appropriately.  Wrap long lines.
 814
 815 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
 816
 817         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
 818         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
 819
 820 2007-01-19  Tomas Bily  <tbily@suse.cz>
 821
 822         * cgraphunit.c (cgraph_finalize_function): Updating of pid
 823         * tree-profile.c:
 824         (tree_init_ic_make_global_vars): New function
 825         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
 826         (tree_gen_ic_profiler): New function
 827         (tree_gen_ic_func_profiler): New function
 828         (tree_profiling): Added calling of tree_gen_ic_func_profiler
 829         (tree_profile_hooks): Added hook for indirec/virtual calls
 830         * value-prof.c (tree_find_values_to_profile): New case for
 831         indirect calls
 832         (tree_values_to_profile): Call for determining indirect/virtual
 833         counters
 834         (tree_indirect_call_to_profile): New function
 835         (tree_ic_transform): New function
 836         (tree_ic): New function
 837         (find_func_by_pid): New function
 838         (init_pid_map): New function
 839         (tree_value_profile_transformations): Added check for
 840         indirect/virtual call transformation
 841         * value-prof.h (enum hist_type): New counter type for
 842         indirect/virtual calls
 843         (profile_hooks): Added new hook for profiling indirect/virtual
 844         calls
 845         * profile.c (instrument_values): New case for indirect/virtual
 846         call added
 847         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
 848         (GCOV_COUNTER_V_INDIR): New counter type
 849         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
 850         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
 851         call added
 852         * cgraph.c: Definition of cgraph_max_pid
 853         (cgraph_create_node): Default init of pid attribute
 854         * cgraph.h: Declaration of cgraph_max_pid
 855         (struct cgraph_node): Added pid attribute
 856         * libgcov.c (__gcov_indirect_call_profiler): New function
 857         (__gcov_one_value_profiler_body): New function
 858         (__gcov_one_value_profiler): Body was moved to
 859         __gcov_one_value_profiler_body and calls it
 860
 861 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
 862
 863         * doc/gty.texi (Options): Document the mark_hook option to GTY.
 864         * gengtype.c (write_types_data, write_func_for_structure,
 865         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
 866         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
 867         (walk_type, write_func_for_structure): Generate the mark_hook if
 868         needed.
 869
 870 2007-01-19  Jan Hubicka  <jh@suse.cz>
 871
 872         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
 873         'early' argument take inlining mode argument specifying whether to
 874         inline for size/speeed or all functions; add support for flattening;
 875         improve dumpting.
 876         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
 877
 878 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
 879
 880         PR c++/17947
 881         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
 882         the name of the declared identifier.
 883
 884 2007-01-19  Dirk Mueller  <dmueller@suse.de>
 885
 886         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
 887         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
 888         duplicate evaluation.
 889
 890 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
 891
 892         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
 893         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
 894         to XFMode, use acosxf2 and truncate result to requested mode.
 895         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
 896         (asin<mode>2): Similarly, with asin expanders.
 897         (*fscalexf4_i387): Rename from *fscalexf4.
 898         (expNcorexf3): New expander.
 899         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
 900         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
 901         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
 902         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
 903         patterns for SSE math.
 904         (exp10<mode>2): Similarly, with exp10 expanders.
 905         (exp2<mode>2): Similarly, with exp2 expanders.
 906         (expm1<mode>2): Similarly, with expm1 expanders.
 907         (ldexp<mode>3): Similarly, with ldexp expanders.
 908         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
 909         Use gen_truncxf<mode>2_i387_noop to truncate result.
 910
 911 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
 912
 913         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
 914         for ColdFire targets.
 915
 916 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
 917             Richard Sandiford  <richard@codesourcery.com>
 918
 919         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
 920         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
 921         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
 922         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
 923         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
 924         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
 925
 926 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
 927
 928         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
 929         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
 930         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
 931         (*adddi_dilshr32_cf): New pattern.
 932         (adddi3, subdi3): Remove first alternatives.
 933
 934 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
 935
 936         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
 937         implemented using move.l, do not use its cc result for floating-point
 938         comparisons.
 939
 940 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
 941
 942         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
 943         from accepting 's' constraints if flag_pic.
 944
 945 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
 946
 947         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
 948         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
 949         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
 950         condition from TARGET_68881 to TARGET_HARD_FLOAT.
 951
 952 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
 953
 954         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
 955         ColdFire alternatives.
 956         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
 957         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
 958         architectures that support ff1 instruction.
 959
 960 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
 961             Julian Brown  <julian@codesourcery.com>
 962
 963         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
 964         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
 965         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
 966         Add commentary.
 967
 968 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
 969
 970         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
 971         and return a bool.
 972         (output_move_const_into_data_reg, output_move_simode_const): Delete.
 973         * config/m68k/m68k.c (const_method, const_int_cost): Take a
 974         HOST_WIDE_INT instead of an rtx.
 975         (m68k_rtx_costs): Update call accordingly.
 976         (output_move_const_into_data_reg): Likewise.  Fix formatting.
 977         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
 978         Return a bool.
 979         (output_move_simode_const): Update calls after above changes.
 980         Rework to use automatic variables and predicates like MEM_P.
 981         * config/m68k/m68k.md (pushexthisi_const): Update call to
 982         valid_mov3q_const.
 983
 984 2007-01-19  Dirk Mueller  <dmueller@suse.de>
 985
 986         * tree-ssa-alias.c (perform_var_substitution): Fix typo
 987         in dump_flags test.
 988
 989 2007-01-19  Richard Guenther  <rguenther@suse.de>
 990
 991         * builtins.c (expand_builtin_cexpi): Fall back to expanding
 992         via cexp in case sincos is not available.
 993
 994 2007-01-19  Richard Guenther  <rguenther@suse.de>
 995
 996         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
 997         * defaults.h (TARGET_HAS_SINCOS): Default to off.
 998         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
 999         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1000         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1001         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1002         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1003         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1004
1005 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
1006
1007         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1008         New insn patterns.
1009         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1010         (atan2xf3): Directly generate RTL pattern.
1011         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1012         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1013         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1014         disable patterns for SSE math.
1015         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1016         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
1017         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1018         disable patterns for SSE math.
1019
1020 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
1021
1022         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1023
1024 2007-01-18  Roger Sayle  <roger@eyesopen.com>
1025
1026         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1027         VIEW_CONVERT_EXPR to the same type as it's operand.
1028
1029 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
1030
1031         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1032         configured for long double 128.
1033
1034 2007-01-18  Mike Stump  <mrs@apple.com>
1035
1036         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1037         error.
1038
1039 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
1040
1041         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1042         type correct.
1043
1044 2007-01-18  Jan Hubicka  <jh@suse.cz>
1045
1046         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1047         greater than NUM_VOP_FREE_BUCKETS.
1048
1049 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
1050
1051         * tree-ssa-structalias.c: Update comments.
1052         (ptabitmap_obstack): Removed.
1053         (pta_obstack): New.
1054         (oldpta_obstack): Ditto.
1055         (stats): Add a few members.
1056         (struct variable_info): Remove node, complex, address_taken, and
1057         indirect_target members. Add oldsolution member.
1058         (new_var_info): Do not initialize removed members.
1059         (constraint_expr_type): Remove INCLUDES.
1060         (constraint_graph): Add size, implicit_preds, rep,
1061         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1062         (FIRST_REF_NODE): New macro.
1063         (LAST_REF_NODE): Ditto.
1064         (FIRST_ADDR_NODE): Ditto.
1065         (find): New function.
1066         (unite): Ditto.
1067         (dump_constraint): Do not handle INCLUDES.
1068         (insert_into_complex): Do not insert duplicate constraints.
1069         (condense_varmap_nodes): Renamed and rewritten into ...
1070         (merge_node_constraints): This. Also fix bug in handling of
1071         offseted copy constraints.
1072         (clear_edges_for_node): No longer need to deal with preds at all,
1073         or removing associated preds/succs.
1074         (merge_graph_nodes): Deal with indirect_cycles.
1075         Don't deal with predecessors.
1076         (add_implicit_graph_edge): New function.
1077         (add_pred_graph_edge): Ditto.
1078         (add_graph_edge): Don't deal with predecessors.
1079         (build_constraint_graph): Removed.
1080         (build_pred_graph): New function.
1081         (build_succ_graph): Ditto.
1082         (struct scc_info): Removed in_component. Added roots, dfs, and
1083         node_mapping. Remove visited_index, unification_queue.
1084         (scc_visit): Deal with union-find we do now.
1085         Deal with cycles with REF nodes.
1086         (collapse_nodes): Renamed and rewritten to ...
1087         (unify_nodes): This.
1088         (process_unification_queue): Removed.
1089         (topo_visit): Cleanup
1090         (do_da_constraint): Use find.
1091         (do_sd_constraint): Ditto.
1092         (do_ds_constraint): Ditto.
1093         (do_complex_constraint): Ditto.
1094         (init_scc_info): Update for removed and added members.
1095         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1096         (find_indirect_cycles): This.
1097         (equivalence_class): New variable.
1098         (label_visit): New function.
1099         (perform_variable_substitution): Rewritten.
1100         (free_var_substitution_info): New function.
1101         (find_equivalent_node): Ditto.
1102         (move_complex_constraints): Ditto.
1103         (eliminate_indirect_cycles): Ditto.
1104         (solve_graph): Only propagate changed bits.
1105         Use indirect cycle elimination.
1106         Use find.
1107         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1108         (tree_id_eq): Renamed to ...
1109         (tree_vi_eq): This. Update for member change
1110         (insert_id_for_tree): Renamed and rewritten to ...
1111         (insert_vi_for_tree): This.
1112         (lookup_id_for_tree): Renamed and rewritten to ...
1113         (lookup_vi_for_tree): This.
1114         (get_id_for_tree): Renamed and rewritten to ...
1115         (get_vi_for_tree): Ditto.
1116         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1117         (process_constraint): Don't handle INCLUDES.
1118         Remove special ADDRESSOF case.
1119         (find_func_aliases): Rewrite to use vi functions instead of id
1120         ones.
1121         (create_function_info_for): Ditto.
1122         (create_variable_info_for): Ditto.
1123         (intra_create_variable_infos): Ditto.
1124         (merge_smts_into): Ditto.
1125         (find_what_p_points_to): Ditto.
1126         (init_base_vars): Ditto.
1127         (init_alias_vars): Ditto.
1128         (remove_preds_and_fake_succs): New function.
1129         (dump_sa_points_to_info): Dump new stats.
1130         (dump_solution_for_var): Use find.
1131         (set_used_smts): Fix formatting.
1132         (compute_points_to_sets): Updated for new functions.
1133         (ipa_pta_execute): Ditto.
1134
1135 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1136             Richard Sandiford  <richard@codesourcery.com>
1137
1138         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1139         parallels.
1140         * calls.c (expand_call): If the return value is a PARALLEL,
1141         extract its first member.
1142         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1143         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1144         use FUNCTION_EXTRA_EPILOGUE.
1145         (m68k_function_value): Return a PARALLEL if the return value
1146         is of a pointer type.
1147         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1148         (FUNCTION_EXTRA_EPILOGUE): Remove.
1149         * config/m68k/m68k.md (D0_REG): New constant.
1150
1151 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1152
1153         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1154         output a NOP for empty epilogues.
1155
1156 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1157
1158         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1159         before function.  Extend register save check to include all
1160         registers, not just integer ones.
1161
1162 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1163
1164         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1165         (m68k_use_return_insn): ...this.
1166         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1167         * config/m68k/m68k.c (use_return_insn): Rename to...
1168         (m68k_use_return_insn): ...this.
1169         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1170         of USE_RETURN_INSN.
1171
1172 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1173
1174         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1175         nearest even, fix denormal rounding overflow.
1176
1177 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1178
1179         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1180         f%$smove and f%$move.
1181         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1182         (extendsfdf2_cf): Use fdmove instead of f%&move.
1183         (truncdfsf2_cf): Use fsmove instead of f%$smove.
1184         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1185
1186 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1187
1188         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1189         GPR<-GPR moves.
1190
1191 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1192
1193         * real.h (real_format): Add a canonical_nan_lsbs_set field.
1194         (coldfire_single_format): Declare.
1195         (coldfire_double_format): Likewise.
1196         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1197         of qnan_msb_set to determine the lower bits of a canonical
1198         NaN significand.
1199         (encode_ieee_double): Likewise.
1200         (encode_ieee_quad): Likewise.
1201         (ieee_single_format): Initialize canonical_nan_lsbs_set.
1202         (mips_single_format): Likewise.
1203         (ieee_double_format): Likewise.
1204         (mips_double_format): Likewise.
1205         (ieee_extended_motorola_format): Likewise.
1206         (ieee_extended_intel_96_format): Likewise.
1207         (ieee_extended_intel_128_format): Likewise.
1208         (ieee_extended_intel_96_round_53_format): Likewise.
1209         (ibm_extended_format): Likewise.
1210         (mips_extended_format): Likewise.
1211         (ieee_quad_format): Likewise.
1212         (mips_quad_format): Likewise.
1213         (vax_f_format): Likewise.
1214         (vax_d_format): Likewise.
1215         (vax_g_format): Likewise.
1216         (i370_single_format): Likewise.
1217         (i370_double_format): Likewise.
1218         (decimal_single_format): Likewise.
1219         (decimal_double_format): Likewise.
1220         (decimal_quad_format): Likewise.
1221         (c4x_single_format): Likewise.
1222         (c4x_extended_format): Likewise.
1223         (real_internal_format): Likewise.
1224         (coldfire_single_format): New real_format.
1225         (coldfire_double_format): Likewise.
1226         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1227         canonical_nan_lsbs_set.
1228         (pdp11_d_format): Likewise.
1229         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1230         if TARGET_COLDFIRE_CPU.
1231
1232 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1233
1234         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1235         (output_call, m68k_legitimize_call_address): Declare.
1236         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1237         parenthesees.  Add support for a 'W' constraint.
1238         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1239         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1240         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1241         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1242         variables.
1243         (override_options): Initialize them.  Do not set flag_no_function_cse
1244         for TARGET_ID_SHARED_LIBRARY.
1245         (m68k_output_pic_call): Delete.
1246         (m68k_legitimize_call_address): New function.
1247         (print_operand): Remove the %o prefix.  Handle the %p prefix.
1248         (output_call): New function.
1249         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
1250         target address from the GOT if symbolic jumps are not allowed.
1251         * config/m68k/m68k.md (call, general_operand): Do not set
1252         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
1253         Merge separate flag_pic and !flag_pic define_insns into...
1254         (*call, *call_value): ...these new patterns.  Match the address
1255         rather than the containing MEM and require it to be a call_operand.
1256         Use output_call to generate the asm template.
1257         * config/m68k/predicates.md (const_call_operand): New predicate.
1258         (call_operand): Likewise.
1259
1260 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1261
1262         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1263         (STACK_GROWS_DOWNWARD): Define to 1.
1264         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1265         comments, emphasizing that these values are only defaults.
1266         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1267         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1268
1269 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1270
1271         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1272
1273 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
1274             Richard Sandiford  <richard@codesourcery.com>
1275
1276         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1277         (FP_REGNO_P): New macros.
1278         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1279         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1280         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1281         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1282
1283 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1284
1285         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1286         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1287         to $tmake_file.
1288         * config/m68k/t-floatlib: New file, extracting common code from...
1289         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1290         * config/m68k/t-uclinux: Here.
1291         * config/m68k/fpgnulib.c: Do not compile extendeed precision
1292         routines on ColdFire targets.
1293
1294 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
1295
1296         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1297         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1298         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1299         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1300         explicitly set MOTOROLA to 1.
1301         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1302
1303 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1304             Nathan Sidwell  <nathan@codesourcery.com>
1305
1306         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1307         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1308         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1309         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1310         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1311         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1312         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1313         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1314         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1315         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1316         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1317         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1318         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1319         (DBX_REGISTER_NUMBER): Delete.
1320         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1321         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1322         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1323
1324 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1325
1326         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1327         * config/flat.h: New file.
1328         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1329         OBJECT_FORMAT_FLAT.
1330         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1331         indirect references for -msep-data or -mid-shared-library.
1332         Do not use PC-relative code addresses either.
1333         * config/m68k/m68k.c (override_options): Restrict -fPIC error
1334         to -mpcrel.
1335         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
1336         for shared libraries and crt1.o for executables.  Use crti.o and
1337         crtbegin.o.
1338         (ENDFILE_SPEC): Use crtend.o and crtn.o.
1339         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1340         Do not add -elf2flt or -shared-lib-id options here.
1341         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
1342         Pass -shared-lib-id if -mid-shared-library, taking the library
1343         identifier from -mshared-library-id if given, otherwise
1344         defaulting to 0.
1345         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1346         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1347         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1348         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1349         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1350         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1351         and crtend.o.
1352         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1353         jump sequence for ISA A and ISA A+.
1354         (PICJUMP): Likewise.
1355
1356 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1357
1358         * config.gcc (m68k-*-uclinux*): Base the port on the common
1359         and m68k GNU/Linux files rather than on the generic ELF ones.
1360         * config/m68k/uclinux.h (TARGET_VERSION): Override.
1361         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1362
1363 2007-01-18  Julian Brown  <julian@codesourcery.com>
1364             Richard Sandiford  <richard@codesourcery.com>
1365
1366         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1367         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1368         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1369         before redefining.
1370         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1371         80 unconditionally.
1372         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1373
1374 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
1375
1376         * doc/install.texi: Document m68k-uclinuxoldabi.
1377         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1378         * config/m68k/uclinux-oldabi.h: New file, copied from
1379         config/m68k/uclinux.h.
1380
1381 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
1382
1383         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1384         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1385         (m32r_compute_frame_size): Use unsigned for regno.
1386         (m32r_reload_lr): Use gen_frame_mem.
1387         (pop): New.
1388         (m32r_output_function_epilogue): Don't output the function epilogue
1389         textually here.
1390         (m32r_expand_epilogue): New.
1391         (direct_return): Return false if the function has the interrupt
1392         attribute.
1393         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1394         * config/m32r/m32r.md (epilogue): New expander.
1395         (return_lr, return_rte): New insns.
1396         (return): Make it expander.
1397         (return_normal): New expander.
1398
1399 2007-01-18  Josh Conner  <jconner@apple.com>
1400
1401         PR target/30485
1402         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1403         support for UNLE, UNLT, UNGE, and UNGT.
1404
1405 2007-01-18  Jan Hubicka  <jh@suse.cz>
1406
1407         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1408         * tree-tailcall (add_virtual_phis): Likewise.
1409         (optimize_tail_call): Return TODOs.
1410         (execute_tail_calls): Return TODOs.
1411         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1412         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1413         changed.
1414         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1415         Cleanup cfg using TODO.
1416         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1417         * passes.c (execute_function_todo): When cleanup did something, remove
1418         unused locals.
1419         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1420         (make_edges): Don't cleanup_cfg.
1421
1422 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
1423
1424         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1425         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1426         operate on the top of stack.
1427         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1428         register handling of instructions that output to the second
1429         stack slot.
1430         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1431         (move_for_stack_reg): Special-case check for dead destination
1432         stack slot for constant load of 1.0 inside UNSPEC_TAN.
1433
1434         * config/i386/i386.md (UNSPEC_TAN): New constant.
1435         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1436         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1437         to correctly model move of constant 1.0 to top stack slot.
1438         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1439         (unnamed peephole2 pattern): Remove corresponding peephole2
1440         pattern that optimizes tan insn and loading of constant 1.0.
1441         (tanxf2): Use fptanxf4_i387.
1442         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1443         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
1444         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
1445         disable patterns for SSE math.
1446         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1447         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1448
1449 2007-01-18  Dirk Mueller  <dmueller@suse.de>
1450             Richard Guenther <rguenther@suse.de>
1451
1452         PR diagnostic/8268
1453         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1454         * common.opt (Warray-bounds): Add new warning option.
1455         * c-opts.c (c_common_handle_option): Define -Warray-bounds
1456         if -Wall is given.
1457         * Makefile.in: make tree-vrp.o depend on toplev.h
1458         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1459         is enabled.
1460         (check_array_refs, check_array_bounds, check_array_ref): New.
1461
1462 2007-01-18  Jan Hubicka  <jh@suse.cz>
1463
1464         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1465         (execute_ssa_ccp): Return flags conditionally.
1466         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1467         changed.
1468         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1469
1470 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
1471
1472         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1473         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1474
1475         * Makefile.c (GTFILES): Add cfglayout.h.
1476         * gengtype.c (open_base_files): Likewise.
1477         * cfglayout.c (cfg_layout_function_footer,
1478         cfg_layout_function_header) Reindent to make gengtype happy.
1479         * cfglayout.h (cfg_layout_function_footer,
1480         cfg_layout_function_header): Add GTY(()) marker.
1481
1482         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1483
1484 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1485
1486         * genautomata.c (write_automata): Include xstrerror output in the
1487         error message if writing the DFA description file fails.
1488
1489 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1490
1491         * config/mips/mips-protos.h (mips_output_external): Make it
1492         return void.
1493         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1494         * config/mips/mips.c (irix_output_external_libcall): Likewise.
1495         (extern_list): Likewise.
1496         (extern_head): Likewise.
1497         (TARGET_ASM_FILE_END): Likewise.
1498         (mips_file_end): Likewise.
1499         (mips_output_external): Rewritten.
1500
1501 2007-01-18  Ben Elliston  <bje@au.ibm.com>
1502
1503         * genpreds.c (write_insn_preds_c): Only write out the function
1504         body for regclass_for_constraint if we have register constraints.
1505
1506 2007-01-17  Tom Tromey  <tromey@redhat.com>
1507
1508         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1509         * doc/install.texi (Testing): Use sourceware.org.
1510         (Binaries): Likewise.
1511         (Specific): Likewise.
1512         * doc/contrib.texi (Contributors): Use sourceware.org.
1513
1514 2007-01-17  Anatoly Sokolov <aesok@post.ru>
1515
1516         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1517         (AVR_ENHANCED): Rename to ...
1518         (AVR_HAVE_MUL): ... new.
1519         (avr_enhanced_p): Rename to ...
1520         (avr_have_mul_p): ... new.
1521         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1522         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1523         * config/avr/avr.c (avr_enhanced_p): Rename to ...
1524         (avr_have_mul_p): ... new.
1525         (base_arch_s): Rename 'enhanced' to 'have_mul'.
1526         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1527         'avr_enhanced_p' and 'enhanced'.
1528         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1529         AVR_HAVE_MUL instead of AVR_ENHANCED.
1530         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1531         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1532         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1533         instead of AVR_ENHANCED.
1534         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1535         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1536         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1537
1538 2007-01-17  Ian Lance Taylor  <iant@google.com>
1539
1540         * vec.h (VEC_reserve_exact): Define.
1541         (vec_gc_p_reserve_exact): Declare.
1542         (vec_gc_o_reserve_exact): Declare.
1543         (vec_heap_p_reserve_exact): Declare.
1544         (vec_heap_o_reserve_exact): Declare.
1545         (VEC_OP (T,A,reserve_exact)): New static inline function, three
1546         versions.
1547         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1548         negative parameter.
1549         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1550         (VEC_OP (T,A,copy)) [all versions]: Likewise.
1551         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1552         * vec.c (calculate_allocation): Add exact parameter.  Change all
1553         callers.
1554         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1555         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1556         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1557         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1558         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1559         (vec_heap_p_reserve_exact): New function.
1560         (vec_heap_o_reserve_exact): New function.
1561
1562 2007-01-17  Jan Hubicka  <jh@suse.cz>
1563
1564         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1565
1566 2007-01-17  Eric Christopher  <echristo@apple.com>
1567
1568         * config.gcc: Support core2 processor.
1569
1570 2007-01-16  Jan Hubicka  <jh@suse.cz>
1571
1572         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1573         the names, just unlink the chain so we don't crash on dangling pointers
1574         to dead SSA names.
1575
1576 2007-01-16  Jan Hubicka  <jh@suse.cz>
1577
1578         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1579         * tree-pass.h: Reorder to make IPA passes appear toegher.
1580         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1581         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1582         parameters, do not call early inliner.
1583         * ipa-inline.c: Update comments.  Include tree-flow.h
1584         (cgraph_decide_inlining): Do not compute inlining parameters.
1585         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1586         be called with function context set up.
1587         (pass_ipa_inline): Remove unreachable functions before pass.
1588         (cgraph_early_inlining): Simplify assuming to be called from the
1589         PM as local pass.
1590         (pass_early_inline): New pass.
1591         (cgraph_gate_ipa_early_inlining): New gate.
1592         (pass_ipa_early_inline): Turn into simple wrapper.
1593         (compute_inline_parameters): New function.
1594         (gate_inline_passes): New gate.
1595         (pass_inline_parameters): New pass.
1596         (apply_inline): Move here from tree-optimize.c
1597         (pass_apply_inline): New pass.
1598         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1599         transforming.
1600         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1601         doing them by hand.
1602         (tree_function_versioning): Do not allocate dummy struct function.
1603         * tree-inline.h (optimize_inline_calls): Update prototype.
1604         * tree-optimize.c (execute_fixup_cfg): Export.
1605         (pass_fixup_cfg): Remove
1606         (tree_rest_of_compilation): Do not apply inlines.
1607         * tree-flow.h (execute_fixup_cfg): Declare.
1608         * Makefile.in (gt-passes.c): New.
1609         * passes.c: Include gt-passes.h
1610         (init_optimization_passes): New passes.
1611         (nnodes, order): New static vars.
1612         (do_per_function_toporder): New function.
1613         (execute_one_pass): Dump current pass here.
1614         (execute_ipa_pass_list): Don't dump current pass here.
1615
1616 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
1617
1618         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1619         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1620
1621 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
1622
1623         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1624         (strong_alias): Define.
1625         (__gcc_qmul): Provide non-FMA for soft-float.
1626         (__gcc_qdiv): Same.
1627         (__gcc_qneg): New.
1628         (__gcc_qeq): New.
1629         (__gcc_qle): New.
1630         (__gcc_qge): New.
1631         (__gcc_qunord): New.
1632         (__gcc_stoq): New.
1633         (__gcc_dtoq): New.
1634         (__gcc_qtos): New.
1635         (__gcc_qtod): New.
1636         (__gcc_qtoi): New.
1637         (__gcc_qtou): New.
1638         (__gcc_itoq): New.
1639         (__gcc_utoq): New.
1640         (fmsub): New.
1641         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1642         soft-float functions.
1643         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1644         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1645         about long double soft float.
1646
1647 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
1648             Tehila Meyzels  <tehila@il.ibm.com>
1649
1650         * tree-vectorizer.h (is_pattern_stmt_p): New.
1651         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1652         formatting (tabs instead of spaces). Cleanup and clarify setting
1653         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1654         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1655         (vectorizable_type_demotion): Check that types are integral.
1656         (vectorizable_type_promotion): Likewise.
1657         (vectorizable_store): Fix typo.  Eliminate new-line at end of
1658         comments.
1659
1660 2007-01-16  Jan Hubicka  <jh@suse.cz>
1661
1662         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1663         cgraph edges too.
1664
1665 2007-01-15  Eric Christopher  <echristo@apple.com>
1666
1667         * ifcvt.c: Include vec.h, vecprim.h.
1668         (check_cond_move_block): New argument regs.
1669         Reorganize. Add registers used to regs.
1670         (cond_move_process_if_block): Use regs set above as
1671         loop bounds.
1672
1673 2007-01-15  Eric Christopher  <echristo@apple.com>
1674
1675         * config/darwin.h: Update copyright.
1676         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1677         (LINK_COMMAND_SPEC): Add -u.
1678         (LINK_SPEC): Fix umbrella for above.
1679
1680 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
1681
1682         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1683         glibc CVS.
1684
1685 2007-01-15  Tom Tromey  <tromey@redhat.com>
1686
1687         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1688         * doc/install.texi (Testing): Don't mention jacks.
1689         (Configuration): Document --enable-java-maintainer-mode.  Move
1690         --with-java-home to libgcj-specific section.  Document
1691         --with-ecj-jar.
1692         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1693
1694 2007-01-15  Jan Hubicka  <jh@suse.cz>
1695
1696         * tree-ssa-dce.c (DCE_TODOs): New.
1697         (propagate_necessity): Return if something changed.
1698         (eliminate_unnecessary_stmts): Likewise.
1699         (perform_tree_ssa_dce): Return TODO flags when needed.
1700         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1701
1702 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
1703
1704         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1705         (fyl2x_extend<mode>xf3_i387): New insn pattern.
1706         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1707         patterns using X87MODEF12 mode macro.  Extend operand 1
1708         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1709         (log10<mode>2): Ditto.
1710         (log2<mode>2): Ditto.
1711         (log1p<mode>2): Ditto.
1712         (logb<mode>2): Ditto.
1713         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1714         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1715         (*fxtractxf3_i387): Rename from *fxtractxf3.
1716         (fxtract_extend<mode>xf3_i387): New insn pattern.
1717         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1718         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1719         and gen_fyl2xxf3_i387().
1720
1721 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
1722
1723         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1724         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1725         get_loop_exit_edges.  Changed return type to bool.
1726         * cfgloop.h (fix_loop_placement): Declaration removed.
1727
1728 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1729
1730         * param.h (MIN_VECT_LOOP_BOUND): New.
1731         * params.def (MIN_VECT_LOOP_BOUND): New.
1732         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1733         argument - minimum threshold for number of iterations.
1734         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1735         argument to declaration.
1736         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1737         MIN_VECT_LOOP_BOUND.
1738         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1739         slpeel_tree_peel_loop_to_edge with additional argument.
1740         (vect_do_peeling_for_alignment): Likewise.
1741         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1742
1743 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1744
1745         PR target/30413
1746         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1747         operands of size 1.
1748
1749 2007-01-14  Jan Hubicka  <jh@suse.cz>
1750
1751         * tree-dfa.c (remove_referenced_var): New function.
1752         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1753         prune referenced vars list too.
1754         * tree-flow.h (remove_referenced_var): Declare.
1755
1756 2007-01-14  Jan Hubicka  <jh@suse.cz>
1757
1758         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1759         separately.
1760         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1761         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1762         * except.h (verify_eh_throw_table_statements): Kill prototype.
1763         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1764         into call exrepssions.
1765         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1766         eh trhow verifier.
1767         * tree-cfg.c: Include pointer-set.h.
1768         (verify_node_sharing): Work on pointer set.
1769         (verify_eh_throw_stmt_node): New.
1770         (verify_stmts): Use pointers sets, verify throw_stmt.
1771
1772 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1773
1774         * ipa-reference.c (analyze_function): Consider also addresses taken
1775         in phi nodes.
1776
1777 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1778
1779         * c-typeck.c (null_pointer_constant_p): Replace use of
1780         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1781         (build_c_cast): Likewise.
1782
1783 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1784
1785         * tree.h (force_fit_type_double): Remove unused final argument.
1786         * c-common.c (constant_expression_warning): Replace use of
1787         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1788         (convert_and_check): Likewise.
1789         (shorten_compare): Update call to force_fit_type_double.
1790         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1791         * convert.c (convert_to_pointer): Update call to
1792         force_fit_type_double.
1793         * fold-const.c (force_fit_type_double): Remove overflowed_const
1794         argument.
1795         (int_const_binop, fold_convert_const_int_from_int,
1796         fold_convert_const_int_from_real, fold_div_compare,
1797         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1798         fold_abs_const, fold_not_const): Remove the final argument from
1799         calls to force_fit_type_double.
1800
1801 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1802
1803         * configure.ac: Set insn to "nop" for spu-*-* also.
1804         * configure: Regenerate.
1805
1806 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
1807
1808         * builtins.def : Add BUILT_IN_FREE.
1809
1810 2007-01-12  Jan Hubicka  <jh@suse.cz>
1811
1812         PR tree-optimization/30443
1813         * tree-inline.c (tree_function_versioning): Do not optimize when
1814         cloning for inlining.
1815
1816 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
1817
1818         * doc/loop.texi: Document recording of loop exits.
1819         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1820         (update_single_exits_after_duplication,
1821         update_single_exit_for_duplicated_loop,
1822         update_single_exit_for_duplicated_loops): Removed.
1823         (duplicate_loop_to_header_edge): Do not call
1824         update_single_exits_after_duplication and
1825         update_single_exit_for_duplicated_loops.
1826         (loop_version): Do not update single_exit information.
1827         (fix_loop_structure): Use record_loop_exits instead of
1828         mark_single_exit_loops.
1829         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1830         the lists of loop exits.
1831         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1832         split_edge, merge_blocks): Update the lists of loop exits.
1833         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1834         loop_optimizer_init.
1835         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1836         of mark_single_exit_loops.
1837         (loop_optimizer_finalize): Call release_recorded_exits.
1838         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1839         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1840         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1841         update single exit information.
1842         * lambda-code.c (perfect_nestify): Ditto.
1843         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1844         (mark_single_exit_loops): Removed.
1845         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1846         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1847         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1848         functions.
1849         (get_loop_exit_edges, single_exit): Use recorded exit lists.
1850         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1851         (verify_loop_structure): Verify consistency of the exit lists.
1852         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
1853         (set_single_exit): Removed.
1854         * cfgloop.h (struct loop_exit): New function.
1855         (struct loop): single_exit_ field replaced by exits field.
1856         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1857         (struct loops): Added exits hash.
1858         (mark_single_exit_loops, set_single_exit): Declaration removed.
1859         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1860
1861 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1862
1863         * doc/invoke.texi: Avoid use of @headitem.
1864
1865 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1866
1867         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1868
1869 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1870
1871         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1872         list of targets.
1873
1874 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1875             Richard Sandiford  <richard@codesourcery.com>
1876
1877         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1878         * config/m68k/m68k.opt: Resort options.
1879         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1880
1881 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1882             Richard Sandiford  <richard@codesourcery.com>
1883
1884         * doc/invoke.texi: Document the macros that are defined by
1885         m68k's -mtune and -mhard-float options.
1886         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1887         (m68k_cpp_cpu_family): Likewise.
1888         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1889         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
1890         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
1891         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1892         (m68k_cpp_cpu_family): Likewise.
1893
1894 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1895
1896         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1897         macros besides mc68000 as tuning macros.  Use a switch statement
1898         to set them and mcpu32.
1899
1900 2007-01-12  Julian Brown  <julian@codesourcery.com>
1901
1902         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1903         (TARGET_68040_ONLY): Rename to...
1904         (TARGET_68040): ...this.
1905         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1906         * config/m68k/m68k.md: Likewise.
1907
1908 2007-01-12  Julian Brown  <julian@codesourcery.com>
1909             Nathan Sidwell  <nathan@codesourcery.com>
1910             Richard Sandiford  <richard@codesourcery.com>
1911
1912         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1913         the -mcpu= argument associated with the --with-cpu setting.
1914         Define M68K_DEFAULT_TUNE to the default -mtune= option,
1915         if different from the one implied by the -mcpu setting.
1916         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1917         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
1918         * doc/invoke.texi: Mention ColdFire in the introduction to the
1919         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
1920         -mno-div and -mhard-float options.  Make -m68881 a synonym for
1921         -mhard-float.  Document the previously-undocumented -m5206e,
1922         -m528x, -m5307 and -m5407 options.  Tweak the existing option
1923         documentation for consistency.
1924         * doc/install.texi: Mention new --with-cpu arguments.
1925         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1926         default CPU if neither -mcpu nor -march are specified.
1927         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1928         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1929         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
1930         using m68k_tune to decide between families that implement the
1931         same ISA.  Use m68k_tune to set __mcfv4e__.
1932         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1933         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1934         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1935         (FL_ISA_C, FL_ISA_MMU): New macros.
1936         (MASK_COLDFIRE): Delete.
1937         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1938         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1939         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1940         (TARGET_HARD_FLOAT): Do not define here.
1941         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1942         (TUNE_68000): New macro.
1943         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1944         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1945         (TUNE_CFV2): Redefine in terms of m68k_tune.
1946         (uarch_type, target_device, fpu_type): New enums.
1947         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1948         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1949         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1950         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1951         (FL_FOR_isa_c): New macros.
1952         (m68k_isa): New enum.
1953         (m68k_target_selection): New structure.
1954         (all_devices, all_isas, all_microarchs): New tables.
1955         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1956         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1957         (MASK_ALL_CPU_BITS): Delete.
1958         (m68k_find_selection): New function.
1959         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1960         Map the legacy target options to a combination of the new ones.
1961         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1962         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
1963         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1964         settings.
1965         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1966         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1967         properties.
1968         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1969         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1970         * config/m68k/m68k-devices.def: New file.
1971
1972 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1973             Nathan Sidwell  <nathan@codesourcery.com>
1974
1975         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1976         (SUBTARGET_EXTRA_SPECS): New macros.
1977         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1978         use %(asm_cpu_spec) instead.
1979         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1980         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1981         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1982         (EXTRA_SPECS): Rename to...
1983         (SUBTARGET_EXTRA_SPECS): ...this.
1984
1985 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1986             Richard Sandiford  <richard@codesourcery.com>
1987             Julian Brown  <julian@codesourcery.com>
1988
1989         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1990         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1991         to the configuration's default CPU.
1992         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1993         Remove default masks.
1994         (m680[012]0-*-*): Set the default with_cpu to the first part of
1995         the target name.
1996         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1997         (m68k*-*-linux): Extend the --with-cpu handling to...
1998         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
1999         and m68010.  Don't set target_cpu_default2.
2000         * doc/install.texi: Document --with-cpu for m68k.
2001         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2002         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2003         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2004         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2005         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2006         (ASM_SPEC): Remove use of %(asm_cpu_default).
2007         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2008         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2009         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
2010         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2011         __HAVE_FPU__ if TARGET_HARD_FLOAT.
2012         (TARGET_DEFAULT): Delete.
2013         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2014         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2015         (CPP_CPU_SPEC): Delete.
2016         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2017         the appropriate string.
2018         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2019         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2020         (ASM_SPEC): Don't use %(asm_default_spec).
2021         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2022         TARGET_DEFAULT and add MASK_68881.
2023         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2024
2025 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2026
2027         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2028         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2029         MASK_68010 alongside MASK_68020.
2030         * doc/invoke.texi: Document -m68010.
2031         * config/m68k/m68k.opt (m68010): New.
2032         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2033         if TUNE_68010.
2034         (TUNE_68010): New macro.
2035         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2036         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2037         (M68K_CPU_m68332): Add MASK_68010.
2038         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2039         fallback definition.
2040         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2041         defines.
2042         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2043         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
2044         to all entries that use MASK_68020.
2045         (output_move_simode_const, output_move_himode, output_move_qimode)
2046         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2047         instead of TARGET_68020 to select clr behavior.  Remove comment
2048         about there being no TARGET_68010.
2049         * config/m68k/m68k.md: Likewise throughout.
2050
2051 2007-01-12  Julian Brown  <julian@codesourcery.com>
2052
2053         * config/m68k/m68k.h (TARGET_ISAB): New macro.
2054         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2055         * config/m68k/m68k.md: Likewise.
2056
2057 2007-01-12  Julian Brown  <julian@codesourcery.com>
2058
2059         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2060         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2061
2062 2007-01-12  Julian Brown  <julian@codesourcery.com>
2063
2064         * config/m68k/m68k.h (TUNE_68040_60): New macro.
2065         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2066         * config/m68k/m68k.md: Likewise.
2067
2068 2007-01-12  Julian Brown  <julian@codesourcery.com>
2069             Richard Sandiford  <richard@codesourcery.com>
2070
2071         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2072         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2073         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2074         TARGET_CPU32.
2075         (TARGET_CPU32): Rename to...
2076         (TUNE_CPU32): ...this.
2077         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2078         (TUNE_CFV2): New macros.
2079         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2080         remove conditions that are implied by TARGET_68020.
2081         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2082         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2083         (m68k_output_function_epilogue): Likewise.
2084         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
2085         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
2086         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2087         non-68000 timings.  Refactor multiplication and division costs.
2088         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2089         TUNE_CPU32 instead of TARGET_CPU32.
2090         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2091         and TUNE_68060 instead of TARGET_68060.
2092         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2093         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2094         TARGET_CPU32.
2095         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2096         && !TARGET_COLDFIRE" to choose between moveq and clr.
2097         Likewise in the unnamed movsf pattern.
2098         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2099         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
2100         ashiftrt pattern.
2101
2102 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
2103
2104         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2105         of tabbing before backslashes.
2106
2107 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2108
2109         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2110         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2111         (ASM_OUTPUT_INTERNAL_LABEL): Define.
2112
2113 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2114
2115         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2116         find_interesting_uses_cond.
2117         (find_interesting_uses_cond): Use extract_cond_operands.
2118         (rewrite_use_compare): Use extract_cond_operands and
2119         force_gimple_operand_bsi.  Do not call update_stmt.
2120         (determine_use_iv_cost_condition): Use extract_cond_operands.
2121         Return cheaper of using original bound and changing the exit bound.
2122
2123 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
2124
2125         PR tree-optimization/29516
2126         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2127         most_expensive_mult_to_index, addr_to_parts,
2128         create_mem_ref, maybe_fold_tmr): Make the type of
2129         fields of TARGET_MEM_REF sizetype.
2130         (move_fixed_address_to_symbol, move_pointer_to_base):
2131         New functions.
2132         * tree.def (TARGET_MEM_REF): Add comment on types of
2133         the operands.
2134
2135 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
2136
2137         * c-common.c (vector_types_convertible_p): Treat opaque types as
2138         always convertible if they have the same size, but not otherwise.
2139
2140 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
2141
2142         * ifcvt.c (struct noce_if_info): Add comments to the fields.
2143         Remove the b_unconditional field.
2144         (noce_try_sign_mask): Do not look at b_unconditional.
2145         (noce_process_if_block): Do not use merge_if_blocks.  Update
2146         the CFG here.  Do not set b_unconditional.
2147         (cond_move_process_if_block): Likewise.
2148         (find_cond_trap): Likewise.
2149         (check_cond_move_block): Require simple jump insns at the end
2150         of the basic block.
2151
2152 2007-01-11  Jan Hubicka  <jh@suse.cz>
2153
2154         PR tree-optimization/1046
2155         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2156         when alias info is not ready.
2157         (pass_tail_recursion): Do not require aliasing.
2158         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2159         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2160         * tree-ssa-copy.c (pass_copy_prop): Likewise.
2161         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2162         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2163         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2164         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2165         before inlining.
2166         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2167         When aliasing is not build, mark statement as volatile.
2168
2169 2007-01-11  Tom Tromey  <tromey@redhat.com>
2170
2171         PR preprocessor/15185, PR preprocessor/20989:
2172         * doc/cppopts.texi <-MT>: Update description of algorithm for
2173         computing default target.
2174         <-M, -MD>: Reword "basename" text.
2175
2176 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2177
2178         * builtins.c (expand_builtin_pow, expand_builtin_powi,
2179         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2180         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2181         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2182         fold_builtin_bswap, real_constp, fold_builtin_pow,
2183         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2184         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2185         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2186         * convert.c (convert_to_pointer): Likewise.
2187         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2188         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2189         const_binop, fold_convert_const_int_from_int,
2190         fold_convert_const_int_from_real,
2191         fold_convert_const_real_from_real, sign_bit_p,
2192         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2193         fold_sign_changed_comparison, fold_unary, fold_comparison,
2194         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2195         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2196         * print-tree.c (print_node_brief, print_node): Likewise.
2197         * stor-layout.c (place_field, layout_type): Likewise.
2198         * tree-chrec.c (keep_cast): Likewise.
2199         * tree.c (build_vector, build_real, build_real_from_int_cst,
2200         build_complex): Likewise.
2201
2202 2007-01-11  Roger Sayle  <roger@eyesopen.com>
2203
2204         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
2205         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2206
2207 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
2208
2209         * configure.ac (strict1_warn): Rename to strict_warn.
2210         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2211         cc_set_by_configure, quoted_cc_set_by_configure,
2212         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2213         all_boot_languages, all_stagestuff): Remove.
2214         (target_list): Remove bootstrap targets.
2215         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2216         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2217         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2218         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2219         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2220         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2221         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2222         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2223         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2224         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2225         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2226         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2227         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2228         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2229         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2230         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2231         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2232         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2233         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2234         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2235         (mostlyclean): Adjust.
2236         (clean, distclean): Don't mention bootstrap stuff.
2237         * configure: Regenerate.
2238         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2239         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2240         treelang/config-lang.in (stagestuff): Remove.
2241         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2242         stageprofile, stagefeedback, stagestuff): Remove mention.
2243
2244 2007-01-11  Nick Clifton  <nickc@redhat.com>
2245
2246         * config/mcore/predicates.md (mcore_general_movesrc_operand):
2247         Accept CONSTs.
2248         (mcore_general_movdst_operand): Do not accept CONST_INTs.
2249         (mcore_arith_K_S_operand): Run the test for the S constraint not
2250         the test for the M constraint.
2251         (mcore_addsub_operand): Do not accept integer values that are
2252         larger than 32 bits.
2253         * config/mcore/mcore.md: Remove unused constraints from split.
2254         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2255         (addsi3): Likewise.
2256         (allocate_stack): Likewise.
2257         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2258         operands to 32 bits.
2259         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2260         INTVAL.
2261         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2262         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2263         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2264         mcore_output_bclri, mcore_output_andn, output_inline_const,
2265         mcore_output_move, mcore_output_movedouble): Likewise.
2266         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2267         (output_inline_const): Likewise.
2268         (output_inline_const): Fix format strings used in sprintf
2269         statements.
2270         * config/mcore/mcore-protos.h: Update prototypes for changed
2271         functions in mcore.c.
2272         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2273         HOST_WIDE_INT and not int.
2274         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2275         CONST_OK_FOR_N): Likewise.
2276         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2277         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2278         an INTVAL.
2279
2280 2007-01-10  Jan Hubicka  <jh@suse.cz>
2281
2282         * tree-vrp.c (remove_range_assertions): Release defs.
2283         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2284         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2285
2286 2007-01-10  Paul Brook  <paul@codesourcery.com>
2287
2288         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2289         * config/arm/arm.md (mulsi3subsi): New insn.
2290
2291 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
2292
2293         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2294         (tree_transform_and_unroll_loop): New.
2295         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2296         Declare.
2297
2298 2007-01-10  Robert Kennedy <jimbob@google.com>
2299
2300         * fold-const.c (fold_comparison): Fold comparisons like (x *
2301         1000 < 0) to (x < 0).
2302
2303 2007-01-10  Ian Lance Taylor  <iant@google.com>
2304
2305         * tree-pretty-print.c (dump_generic_node): Print parentheses when
2306         operands have the same priority.
2307
2308 2007-01-10  Tom Tromey  <tromey@redhat.com>
2309
2310         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2311         (fold_binary): Likewise.
2312         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2313         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2314         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2315         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2316         * langhooks.h (struct lang_hooks): Removed field
2317         'can_use_bit_fields_p'.
2318
2319 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
2320
2321         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2322
2323 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
2324
2325         * function.c (get_last_funcdef_no): New function.
2326         * function.h (get_last_funcdef_no): Declare.
2327         * tree-inline.c (initialize_cfun): Add initialization.
2328         (tree_function_versioning): Cleanup.
2329
2330 2007-01-10  Jan Hubicka  <jh@suse.cz>
2331
2332         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2333         PHIs.
2334
2335 2007-01-10  Sa Liu  <saliu@de.ibm.com>
2336             Ben Elliston  <bje@au.ibm.com>
2337
2338         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2339         for SAVE_NONLOCAL and Pmode for any other save level.
2340         * spu-protos.h (spu_restore_stack_block): Declare.
2341         * spu.md (save_stack_block): Remove.
2342         (restore_stack_block): Call spu_restore_stack_block.
2343         * spu.c (spu_restore_stack_block): New function.
2344         (spu_expand_epilogue): Remove old comment.
2345
2346 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
2347
2348         PR tree-optimization/30322
2349         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2350         (cand_value_at): Return the value as aff_tree.
2351         (may_eliminate_iv): Convert the bound from aff_tree to tree.
2352         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2353         aff_combination_mult): New functions.
2354         (aff_combination_add): Use aff_combination_add_cst.
2355         (aff_combination_convert): Allow conversions to a wider type.
2356         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2357         * tree-affine.h (aff_combination_mult): Declare.
2358
2359 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
2360
2361         * doc/tm.texi: Update documentation to reflect reality of exec
2362         and start file search behaviours. Update copyright year.
2363         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2364         header file directories.
2365
2366 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
2367
2368         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2369         (*cosxf2): Rename to cosxf2_i387.
2370         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
2371         using X87MODEF12 mode macro. Rename patterns to
2372         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
2373         for SSE math.
2374         (*cosdf2, *cossf2): Ditto.
2375         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
2376         to match extended input operands.
2377         (sincos<mode>3): New expander.
2378         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2379         insn patterns and corresponding splitters.
2380
2381 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
2382
2383         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2384         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2385
2386 2007-01-09  Nicolas Pitre  <nico@cam.org>
2387
2388         PR target/30173
2389         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2390
2391 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2392
2393         * target.h (struct gcc_target): New field library_rtti_comdat.
2394         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2395         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2396         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2397         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2398
2399 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
2400
2401         * doc/invoke.texi (Optimize Options): Correct description of -O0.
2402
2403 2007-01-08  Richard Guenther  <rguenther@suse.de>
2404
2405         * tree.h (force_fit_type_double): Export.
2406         (force_fit_type): Remove.
2407         * fold-const.c (force_fit_type_double): New function.
2408         (force_fit_type): Remove.
2409         (int_const_binop): Use it.
2410         (fold_convert_const_int_from_int): Likewise.
2411         (fold_convert_const_int_from_real): Likewise.
2412         (fold_div_compare): Likewise.
2413         (fold_sign_changed_comparison): Likewise.
2414         (fold_unary): Likewise.
2415         (fold_negate_const): Likewise.
2416         (fold_abs_const): Likewise.
2417         (fold_not_const): Likewise.
2418         * c-common.c (shorten_compare): Use force_fit_type_double.
2419         * convert.c (convert_to_pointer): Likewise.
2420
2421 2007-01-08  Richard Guenther  <rguenther@suse.de>
2422
2423         * tree.h (build_int_cst_wide_type): Export.
2424         * tree.c (build_int_cst_wide_type): New function.
2425         (build_int_cst_wide): Fix comment.
2426         * builtins.c (fold_builtin_object_size): Use build_int_cst
2427         to build -1 or 0 of the correct type.  Use fit_double_type
2428         to check for overflow.
2429         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2430         to build the mask.
2431         (decode_field_reference): Likewise.
2432         (all_ones_mask_p): Likewise.
2433         (native_interpret_int): Use build_int_cst_wide_type.
2434         (fold_binary): Use build_int_cst_type to build an all-ones
2435         value.
2436         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2437
2438 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
2439
2440         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2441
2442 2007-01-09  Ben Elliston  <bje@au.ibm.com>
2443
2444         * genautomata.c (STATS_OPTION): New option.
2445         (stats_flag): New flag.
2446         (gen_automata_option): Handle it.
2447         (initiate_automaton_gen): Ditto.
2448         (write_automata): Output statistics only if stats_flag is
2449         set. Likewise, output time statistics only if time_flag is set.
2450         * doc/md.texi (Processor pipeline description): Document new flag.
2451
2452 2007-01-08  Richard Guenther  <rguenther@suse.de>
2453
2454         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2455         * tree.c (build_int_cst_type): Likewise.
2456         (size_in_bytes): Don't call force_fit_type on the result.
2457         (int_fits_type_p): Use fit_double_type.
2458         * fold-const.c (fit_double_type): New function.
2459         (force_fit_type): Use it.
2460         * tree.h (fit_double_type): Export.
2461
2462 2007-01-08  Jan Hubicka  <jh@suse.cz>
2463
2464         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2465         * ipa.c (function_and_variable_visibility): Fix return type.
2466
2467 2007-01-08  Richard Guenther  <rguenther@suse.de>
2468
2469         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2470         of offset to build the index.
2471         * tree-pretty-print.c (dump_generic_node): Don't build negated
2472         const just for printing.
2473         * c-pretty-print.c (pp_c_integer_constant): Likewise.
2474         * builtins.c (fold_builtin_int_roundingfn): Check if result
2475         fits the type by using force_fit_type and comparing the result.
2476         * predict.c (predict_loops): Use compare_tree_int for comparison.
2477         * tree.c (build_int_cst): Fall back to integer_type_node for
2478         NULL_TREE type.
2479         (build_int_cst_wide): Assert type is non-null.
2480
2481 2007-01-08  Roberto Costa  <roberto.costa@st.com>
2482
2483         * tree-vrp.c (extract_range_from_cond_expr): New.
2484         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2485         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2486         as expressions.
2487         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2488         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2489         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2490         used as expressions.
2491         * tree-object-size.c (cond_expr_object_size): New.
2492         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2493
2494 2007-01-08  Jan Hubicka  <jh@suse.cz>
2495
2496         * tree-ssa-forwprop.c (forward_propagate_into_cond,
2497         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2498         statement.
2499
2500 2007-01-08  Richard Guenther  <rguenther@suse.de>
2501
2502         PR tree-optimization/23603
2503         * tree-vrp.c (set_value_range_to_truthvalue): New function.
2504         (extract_range_from_binary): Fall back to truthvalue instead of
2505         varying for TRUTH_*_EXPR.
2506         (extract_range_from_comparison): Fall back to truthvalue instead of
2507         varying.
2508         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2509         if all visited PHI values were constant.
2510
2511 2007-01-08  Jan Hubicka  <jh@suse.cz>
2512
2513         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2514         (cgraph_analyze_function): break out from ...
2515         (cgraph_finalize_compilation_unit): ... here.
2516         (cgraph_expand_function): Remove forgoten commented out line.
2517         (cgraph_optimize): Analyze functions.
2518
2519 2007-01-08  Jan Hubicka  <jh@suse.cz>
2520
2521         * tree-pas.h (TODO_remove_function): New flag.
2522         (TODO_update*): Renumber.
2523         (pass_ipa_increase_alignment,
2524         pass_ipa_function_and_variable_visibility): New passes.
2525         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2526         (cgraph_function_and_variable_visibility): Move to ipa.c
2527         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2528         cgraph_increase_alignment.
2529         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2530         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2531         calling tree-inline.
2532         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2533         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2534         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2535         (gate_increase_alignment): New function.
2536         (pass_ipa_increase_alignment): New pass.
2537         * ipa.c: Inline tree-pass.h and timevar.h
2538         (function_and_variable_visibility): Move here from cgraphunit.c
2539         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2540         * passes.c (init_optimization_passes): Add the two new passes.
2541         (execute_todo): Handle cgraph_remove_functions.
2542
2543 2007-01-08  Nick Clifton  <nickc@redhat.com>
2544
2545         * config/frv/predicates.md (reg_or_0_operand): Accept
2546         CONST_DOUBLEs.
2547
2548 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
2549
2550         * config/bfin/rtems.h, config/bfin/t-rtems: New.
2551         * config.gcc: Add bfin*-rtems*.
2552
2553 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2554
2555         * c.opt: Add -flax-vector-conversions.
2556         * c-typeck.c (convert_for_assignment): Pass flag to
2557         vector_types_convertible_p to allow emission of note.
2558         (digest_init): Likewise.
2559         * c-opts.c: Handle -flax-vector-conversions.
2560         * c-common.c (flag_lax_vector_conversions): New.
2561         (vector_types_convertible_p): Unless -flax-vector conversions
2562         has been passed, disallow conversions between vectors with
2563         differing numbers of subparts and/or element types.  If such
2564         a conversion is disallowed, possibly emit a note on the first
2565         occasion only to inform the user of -flax-vector-conversions.
2566         The new last argument specifies this.
2567         * c-common.h (flag_lax_vector_conversions): New.
2568         (vector_types_convertible_p): Add extra argument.
2569         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2570         char_type_node for V*QI type vectors.
2571         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2572         Update to satisfy new typechecking rules.
2573         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2574         C and C++ variants.
2575         * doc/invoke.texi (C Dialect Options): Document
2576         -flax-vector-conversions.
2577
2578 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
2579
2580         PR tree-optimization/29877
2581         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2582         a register variable on the RHS to not be replaceable.
2583
2584 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
2585         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2586         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2587         const_simm15, const_pow2, const_npow2): Added.
2588         * config/score/misc.md (insv, extv, extzv, movmemsi,
2589         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2590         * config/score/score.c (score_address_cost, score_select_cc_mode):
2591         Added.
2592         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2593         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2594         Update score_print_operand.
2595         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2596         Adjust register allocate order and update some macro define.
2597         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2598         mdx_block_move_straight, mdx_block_move_loop_head,
2599         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2600         mdx_block_move): Added.
2601         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2602         mdp_limm): Updated and fix some bug and typo.
2603         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2604         (movsf, movdf, doloop_end): Added.
2605
2606 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
2607
2608         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2609         config/arm/thumb2.md: Fix comment typos.
2610         * doc/extend.texi: Fix a typo.
2611
2612 2007-01-07  Eric Christopher  <echristo@apple.com>
2613
2614         * configure.ac: Check for __stack_chk_fail for darwin.
2615         * configure: Regenerate.
2616
2617 2007-01-07  Richard Guenther  <rguenther@suse.de>
2618
2619         * tree-vrp.c (extract_range_from_assert): CSE calls to
2620         compare_values where possible.
2621         (extract_range_from_unary_expr): Likewise.
2622
2623 2007-01-07  Anatoly Sokolov <aesok@post.ru>
2624
2625         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2626         * config/avr/avr.c (call_insn_operand): Delete function.
2627         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2628         instead of register number. Use predicates.
2629         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2630         call_insn_operand): Add.
2631
2632 2007-01-06  Jan Hubicka  <jh@suse.cz>
2633
2634         * tree-pass.h (pass_build_cgraph_edges): Declare.
2635         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2636         (visited_nodes): Remove.
2637         (cgraph_create_edges): Move to cgraphbuild.c; rename to
2638         build_cgrpah_edges; make visited_nodes local.
2639         (cgraph_process_new_functions): DO not call initialize_inline_failed.
2640         (record_references_in_initializer): Move to cgraphbuild.c
2641         (initialize_inline_failed, rebuild_cgraph_edges,
2642         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2643         (verify_cgraph_node): Make visited_nodes local.
2644         (cgraph_analyze_function): Do not call cgraph_create_edges and
2645         initialize_inline_failed.
2646         (cgraph_expand_function): Do not call cgraph_lower_function;
2647         assert that function is already lowered.
2648         * Makefile.in (cgraphbuild.o): New.
2649         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2650         at the end of lowering passes.
2651
2652 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
2653
2654         * ifcvt.c (cond_move_convert_if_block): New function, code
2655         factored out from...
2656         (cond_move_process_if_block): ...here.  Call the new function
2657         on the THEN and ELSE blocks.
2658         (merge_if_block): Do not copy global_live_at_end, merge_blocks
2659         already takes care of this.
2660
2661 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2662
2663         PR c/19978
2664         * tree.h (TREE_OVERFLOW_P): New.
2665         * c-typeck.c (parser_build_unary_op): Warn only if result
2666         overflowed and operands did not.
2667         (parser_build_binary_op): Likewise.
2668         (convert_for_assignment): Remove redundant overflow_warning.
2669         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2670
2671 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2672
2673         * c-typeck.c (store_init_value): Split over two lines to follow
2674         the GNU coding style.
2675
2676 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
2677
2678         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2679         __GXX_EXPERIMENTAL_CXX0X__.
2680         * doc/cpp.texi: Same.
2681
2682 2007-01-05  Richard Guenther  <rguenther@suse.de>
2683
2684         PR middle-end/27826
2685         * tree.c (get_narrower): Do not construct COMPONENT_REFs
2686         with mismatched types.  Instead explicitly build a
2687         conversion NOP_EXPR.
2688
2689 2007-01-05  Ian Lance Taylor  <iant@google.com>
2690
2691         * c-common.c (decl_with_nonnull_addr_p): New function.
2692         (c_common_truthvalue_conversion): Call it.
2693         * c-typeck.c (build_binary_op): Likewise.
2694         * c-common.h (decl_with_nonnull_addr_p): Declare.
2695
2696 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
2697
2698         PR c/30360
2699         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2700         denom against 0.0.
2701
2702 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
2703
2704         * doc/install.texi (Final install): Document the fact that
2705         the GNAT runtime should not be stripped.
2706
2707 2007-01-04  Jan Hubicka  <jh@suse.cz>
2708
2709         * tree-inline.c (fold_marked_statements): Update operand caches
2710         and EH after folding
2711
2712 2007-01-04  Ian Lance Taylor  <iant@google.com>
2713
2714         * c-common.c (check_function_nonnull): Whitespace fix.
2715
2716 2007-01-04  Jan Hubicka  <jh@suse.cz>
2717
2718         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2719         comitted older version of patch.
2720         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2721         function, celanup cfg and collect garbage.
2722
2723 2007-01-04  Mike Stump  <mrs@apple.com>
2724
2725         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2726         (clean): Likewise.
2727
2728 2007-01-04  Eric Christopher  <echristo@apple.com>
2729
2730         * libgcc2.c (__bswapsi2): Use SItype.
2731         (__bswapdi2): Use DItype.
2732         * libgcc2.h: Update for above.
2733
2734 2007-01-04  Paul Brook  <paul@codesourcery.com>
2735
2736         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2737         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2738         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2739         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2740         umulsi3_highpart): Make conditional on !arm_arch6.
2741         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2742         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2743         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2744         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2745         umulsi3_highpart_v6): New insns.
2746
2747 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2748
2749         * fold-const.c (fold_convert): When casting an expression to void,
2750         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2751         have a type.  Instead of attempting to build a NOP_EXPR, return
2752         these "special" trees directly.
2753
2754 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2755
2756         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2757         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2758         rs6000_cost->fp not 0 as adjustment for outer NEG.
2759
2760 2007-01-04  Jan Hubicka  <jh@suse.cz>
2761
2762         * cgraph.c (cgraph_release_function_body): New function.
2763         (cgraph_remove_node): Use it.
2764         * cgraph.h (cgraph_release_function_body): Declare.
2765         * cgraphunit.c (cgraph_expand_function): Use it.
2766         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2767         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2768         is initialized and while compilation of other function is running.
2769         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2770         statement CFG annotations from here to ...
2771         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2772
2773 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2774
2775         * cfgloop.h (enum li_flags): Make the constants powers of two.
2776
2777 2007-01-04  Jan Hubicka  <jh@suse.cz>
2778
2779         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2780         set.
2781         (fold_marked_statements): New function.
2782         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2783         * tree-inline.h (copy_body_data): Add statements_to_fold.
2784
2785 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2786
2787         * config.gcc: Mention libgcc/config.host.
2788         * Makefile.in: Update comments mentioning libgcc.
2789         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2790         (all.cross, start.encap, rest.encap, rest.cross): Update
2791         dependencies for libgcc move.
2792         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2793         (clean-target, clean-target-libgcc): Delete.
2794         (srcdirify, GCC_EXTRA_PARTS): New macros.
2795         (libgcc-support, libgcc.mvars): New rules.
2796         (distclean): Remove mention of mklibgcc.
2797         (install): Don't reference INSTALL_LIBGCC.
2798         (install-common): Don't reference EXTRA_PARTS.
2799         (install-libgcc, install-multilib): Delete rules.
2800         * mklibgcc.in: Delete file.
2801         * doc/configfiles.texi: Don't mention mklibgcc.
2802
2803         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2804         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2805         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2806         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2807         (mklibgcc, ldblspecs): Likewise.
2808
2809         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2810         (posixpre.def): Use $(T).
2811         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2812         (SHLIB_SRC, SHLIB_INSTALL): Delete.
2813         (SHLIB_LINK): Make dummy.
2814         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2815
2816         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2817
2818         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2819         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2820         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2821         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2822
2823         * configure.ac (all_outputs): Remove mklibgcc.
2824         * configure: Regenerated.
2825
2826 2007-01-03  Josh Conner  <jconner@apple.com>
2827
2828         PR middle-end/29683
2829         * calls.c (compute_argument_addresses): Set stack and stack_slot
2830         for partial args, too.
2831         (store_one_arg): Use locate.size.constant for the size when
2832         generating a save_area.
2833
2834 2007-01-03  Robert Kennedy <jimbob@google.com>
2835
2836         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2837         whose definitions are deleted due to basic block merging.
2838
2839 2007-01-03  Paul Brook  <paul@codesourcery.com>
2840
2841         PR target/16634
2842         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2843         functions.
2844         (use_return_insn): Return 0 for Thumb interrupt functions.
2845         (print_multi_reg): Add rfe argument for IRQ returns.
2846         (arm_output_epilogue): Pop interrupt return address directly into PC.
2847         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2848
2849 2007-01-03  Paul Brook  <paul@codesourcery.com>
2850
2851         Merge from sourcerygxx-4_1.
2852         * config/arm/thumb2.md: New file.
2853         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2854         Thumb-2.
2855         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2856         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2857         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2858         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2859         tables.
2860         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2861         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2862         comptibility.
2863         * config/arm/ieee754-sf.S: Ditto.
2864         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2865         (thumb1_base_register_rtx_p): ... to this.
2866         (thumb_index_register_rtx_p): Rename...
2867         (thumb1_index_register_rtx_p): ... to this.
2868         (thumb_output_function_prologue): Rename...
2869         (thumb1_output_function_prologue): ... to this.
2870         (thumb_legitimate_address_p): Rename...
2871         (thumb1_legitimate_address_p): ... to this.
2872         (thumb_rtx_costs): Rename...
2873         (thumb1_rtx_costs): ... to this.
2874         (thumb_compute_save_reg_mask): Rename...
2875         (thumb1_compute_save_reg_mask): ... to this.
2876         (thumb_final_prescan_insn): Rename...
2877         (thumb1_final_prescan_insn): ... to this.
2878         (thumb_expand_epilogue): Rename...
2879         (thumb1_expand_epilogue): ... to this.
2880         (arm_unwind_emit_stm): Rename...
2881         (arm_unwind_emit_sequence): ... to this.
2882         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2883         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2884         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2885         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2886         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2887         thumb2_output_casesi): New functions.
2888         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2889         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2890         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2891         THUMB2_WORK_REGS): Define.
2892         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2893         arm_condexec_mask, arm_condexec_masklen)): New variables.
2894         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2895         (arm_override_options): Check new CPU capabilities.
2896         Set new architecture flag variables.
2897         (arm_isr_value): Handle v7m interrupt functions.
2898         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
2899         Thumb-2.
2900         (const_ok_for_arm): Handle Thumb-2 constants.
2901         (arm_gen_constant): Ditto.  Use movw when available.
2902         (arm_function_ok_for_sibcall): Return false for v7m interrupt
2903         functions.
2904         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2905         (thumb_find_work_register, arm_load_pic_register,
2906         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2907         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2908         print_multi_reg, output_mov_long_double_fpa_from_arm,
2909         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2910         output_mov_double_fpa_from_arm, output_move_double,
2911         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2912         output_return_instruction, arm_output_function_prologue,
2913         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2914         arm_output_mi_thunk, thumb_set_return_address): Ditto.
2915         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
2916         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2917         (arithmetic_instr, shift_op): Use arm_shift_nmem.
2918         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
2919         '!' and 'L'.
2920         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2921         (thumb_expand_prologue): Use thumb_set_frame_pointer.
2922         (arm_file_start): Output directive for unified syntax.
2923         (arm_unwind_emit_set): Handle stack alignment instruction.
2924         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2925         Add v6t2, v7, v7a, v7r and v7m.
2926         (RETLDM): Add Thumb-2 code.
2927         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2928         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2929         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2930         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2931         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2932         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2933         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2934         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2935         ADJUST_INSN_LENGTH): Define.
2936         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2937         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2938         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2939         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2940         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2941         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2942         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2943         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2944         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2945         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2946         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2947         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2948         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2949         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2950         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2951         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2952         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2953         cortex-m3.
2954         * config/arm/arm-tune.md: Regenerate.
2955         * config/arm/arm-protos.h: Update prototypes.
2956         * config/arm/vfp.md: Enable patterns for Thumb-2.
2957         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
2958         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2959         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2960         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2961         * config/arm/libunwind.S: Add Thumb-2 code.
2962         * config/arm/constraints.md: Update include Thumb-2.
2963         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2964         * config/arm/ieee754-df.S: Ditto.
2965         * config/arm/bpabi.S: Ditto.
2966         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2967         * config/arm/predicates.md (low_register_operand,
2968         low_reg_or_int_operand, thumb_16bit_operator): New.
2969         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2970         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2971         * config/arm/t-arm-elf: Add armv7 multilib.
2972         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2973         Include thumb2.md.
2974         (UNSPEC_STACK_ALIGN, ce_count): New.
2975         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2976         arm_zero_extendsidi2, arm_zero_extendqidi2): New
2977         insns/expanders.
2978         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2979         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2980         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2981         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2982         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2983         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2984         insns.
2985         * doc/extend.texi: Document ARMv7-M interrupt functions.
2986         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2987
2988 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
2989
2990         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2991         (struct _Unwind_Context): Rename args_size to flags, remove
2992         signal_frame field, add a new args_size field and version field.
2993         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2994         _Unwind_IsExtendedContext): New inline functions.
2995         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2996         Assume by_value array is only present if _Unwind_IsExtendedContext.
2997         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2998         _Unwind_IsSignalFrame.
2999         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3000         (uw_update_context_1): Use _Unwind_SetSignalFrame.
3001         (uw_init_context_1): Initialize context->flags to
3002         EXTENDED_CONTEXT_BIT.
3003         * config/rs6000/linux-unwind.h (frob_update_context): Use
3004         _Unwind_SetSignalFrame.
3005
3006 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3007
3008         PR middle-end/30353
3009         * gimplify.c (gimplify_modify_expr_complex_part): Move below
3010         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
3011         the value.
3012
3013 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
3014
3015         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3016         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3017         RET, AT_SP): Remove.
3018         * config/i386/i386.md (*sse_prologue_save_insn): Use return
3019         instead of RET.
3020
3021         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3022         * cfgbuild.c (find_basic_blocks): Likewise.
3023         * cfgrtl.c (rtl_create_basic_block): Likewise.
3024         * function.c (temp_slots_at_level): Likewise.
3025         * reg-stack.c (stack_regs_mentioned): Likewise.
3026         * regclass.c (allocate_reg_info): Likewise.
3027         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3028         set_bb_for_stmt, move_block_to_fn): Likewise.
3029         * tree-complex.c (tree_lower_complex): Likewise.
3030         * vec.h (VEC_safe_grow_cleared): New.
3031
3032         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3033
3034 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
3035
3036         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3037         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3038         to loop_version.
3039         * cfgloopmanip.c (scale_loop_frequencies): Export.
3040         (loopify): Scale the frequencies by prescribed coefficients.
3041         (set_zero_probability): New function.
3042         (duplicate_loop_to_header_edge): Improve updating of frequencies.
3043         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3044         and frequencies according to arguments.
3045         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3046         correctly.
3047         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3048         * modulo-sched.c (sms_schedule): Set probabilities for entering
3049         versioned loop correctly.
3050         * tree-vect-transform.c (vect_transform_loop): Ditto.
3051         * cfgloop.h (loopify, loop_version): Declaration changed.
3052         (scale_loop_frequencies): Declared.
3053
3054 2007-01-02  Jan Hubicka  <jh@suse.cz>
3055
3056         * cgraph.c: Include tree-flow.h
3057         (cgraph_add_new-function): Handle IPA_SSA mode; execute
3058         early_local_passes.
3059         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3060         * tree-pass.h (pass_all_early_optimizations): Declare.
3061         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3062         early_local_passes.
3063         (cgraph_analyze_function): Do early_local_passes.
3064         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3065         Do not add referenced vars.
3066         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3067         SSA form.
3068         (gate_all_early_local_passes): New gate.
3069         (pass_early_local_passes): Use new gate.
3070         (execute_early_local_optimizations): New functions.
3071         (gate_all_early_optimizations): New gate.
3072         (pass_all_early_optimizations): New pass.
3073         (execute_free_datastructures): Free SSA only when initialized.
3074         (gate_init_datastructures): Init only when optimizing.
3075         (tree_lowering_passes): Do early local passes when called late.
3076         * tree-profile.c (do_tree_profiling): Don't profile functions added
3077         late.
3078         (do_early_tree_profiling, pass_early_tree_profile): Kill.
3079         * tree-cfg.c (update_modified_stmts): Do not update when operands are
3080         not active.
3081         * passes.c (init_optimizations_passes): Reorder so we go into SSA
3082         during early_local_passes.
3083         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3084
3085
3086 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3087
3088         * Makefile.in: Update copyright year.
3089
3090 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
3091
3092         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3093         in $(lang_checks) and check-consistency targets.
3094
3095 2007-01-02  Jan Hubicka  <jh@suse.cz>
3096
3097         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3098         Do not add referenced vars.
3099         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3100         are not active.
3101         * passes.c (init_optimization_passes): Put mudflap_2 after
3102         free_datastructures.
3103
3104 2007-01-02  Jan Hubicka  <jh@suse.cz>
3105
3106         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3107         Set NOTHROW flag on call statements proved to be nothrow.
3108         Update statement of local calls so new pure/const functions are
3109         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3110         goto receivers.
3111         (tree_rest_of_compilation): Register hooks and initialize bitmap
3112         early. Do not set after_inlining flag.
3113
3114 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
3115
3116         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
3117         HOST_BITS_PER_LONGLONG
3118
3119 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3120
3121         PR c/19977
3122         * c-typeck.c (store_init_value): Don't emit pedantic overflow
3123         warning for non-static initializers.
3124
3125 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
3126
3127         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3128         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3129         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3130         Always use set_unique_reg_note to add REG_EQUAL notes.
3131
3132 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3133
3134         Revert:
3135         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3136
3137         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3138         * cfgbuild.c (find_basic_blocks): Likewise.
3139         * cfgrtl.c (rtl_create_basic_block): Likewise.
3140         * function.c (temp_slots_at_level): Likewise.
3141         * reg-stack.c (stack_regs_mentioned): Likewise.
3142         * regclass.c (allocate_reg_info): Likewise.
3143         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3144         set_bb_for_stmt, move_block_to_fn): Likewise.
3145         * tree-complex.c (tree_lower_complex): Likewise.
3146         * vec.h (VEC_safe_grow_cleared): New.
3147
3148 2007-01-02  Ian Lance Taylor  <iant@google.com>
3149
3150         * c-common.c (c_common_truthvalue_conversion): When warning about
3151         using an assignment as a truth value, set TREE_NO_WARNING.
3152
3153 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3154
3155         PR middle-end/7651
3156         * c.opt (Wold-style-declaration): New.
3157         * doc/invoke.texi (C-only Warning Options): New.
3158         (Wold-style-declaration): Document it.
3159         (Wextra): Enabled by -Wextra.
3160         * c-opts.c (c_common_post_options): Enabled by -Wextra.
3161         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3162         -Wold-style-declaration.
3163
3164 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
3165
3166         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3167         * cfgbuild.c (find_basic_blocks): Likewise.
3168         * cfgrtl.c (rtl_create_basic_block): Likewise.
3169         * function.c (temp_slots_at_level): Likewise.
3170         * reg-stack.c (stack_regs_mentioned): Likewise.
3171         * regclass.c (allocate_reg_info): Likewise.
3172         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3173         set_bb_for_stmt, move_block_to_fn): Likewise.
3174         * tree-complex.c (tree_lower_complex): Likewise.
3175         * vec.h (VEC_safe_grow_cleared): New.
3176
3177 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
3178
3179         * c-common.c (c_common_nodes_and_builtins): Since variants of
3180         void_type_node get built before it is given a name, we need to
3181         give those variants the name, too.
3182         (complete_array_type): We need to work with the canonical main
3183         type of the array, from which we will build the qualified version.
3184         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3185         * print-tree.c (print_node): Display canonical type information
3186         for each type.
3187         * stor-layout.c (layout_type): When we don't know the
3188         alignment of a type for which we're building an array, we end up
3189         guessing wrong, so make the type require structural equality.
3190         * tree.c (make_node_stat): When we build a new type, it is its
3191         own canonical type.
3192         (build_type_attribute_qual_variant): When building an attribute
3193         variant, its canonical type is the non-attribute variant. However,
3194         if the attributes are target-dependent and they differ, we need to
3195         use structural equality checks for this type.
3196         (build_qualified_type): A qualified type is not equivalent to its
3197         unqualified variant; set the canonical type appropriately.
3198         (build_distinct_type_copy): When building a distinct type from
3199         another type, the new type is its own canonical type.
3200         (build_variant_type_copy): When building a new type variant, we
3201         assume that it is equivalent to the original type.
3202         (build_pointer_type_for_mode): When building a pointer type, also
3203         build a canonical type pointer.
3204         (build_reference_type_for_mode): When building a reference type,
3205         also build a canonical type reference.
3206         (build_index_type): When we can't hash an index type (e.g.,
3207         because its maximum value is negative), the index type requires
3208         structural equality tests.
3209         (build_array_type): Build the canonical form of an array type.
3210         (build_function_type): Function types require structural equality,
3211         because they contain default arguments, attributes, etc.
3212         (build_method_type_directly): Ditto for method types.
3213         (build_offset_type): Build the canonical offset type.
3214         (build_complex_type): Build the canonical vector type.
3215         (make_vector_type): Build the canonical vector type.
3216         * tree.h (TYPE_CANONICAL): New.
3217         (TYPE_STRUCTURAL_EQUALITY_P): New.
3218         (SET_TYPE_STRUCTURAL_EQUALITY): New.
3219         (struct tree_type): Added "canonical" field.
3220         * params.h (VERIFY_CANONICAL_TYPES): New.
3221         * doc/c-tree.texi (TYPE_CANONICAL): Document.
3222         (TYPE_STRUCTURAL_EQUALITY_P): Document.
3223         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3224         * doc/invoke.texi (verify-canonical-types): Document --param
3225         parameter for verifying canonical types.
3226
3227 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3228
3229         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3230         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3231         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3232         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3233         tm_file.
3234         * config/rs6000/e500.h: New.
3235         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3236         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3237         Remove.
3238         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3239         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3240         TARGET_E500_DOUBLE): Remove.
3241         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3242         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3243         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3244         * config/rs6000/rs6000.c (rs6000_override_options): Use
3245         CHECK_E500_OPTIONS.
3246
3247 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3248
3249         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3250         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3251         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3252         rs6000_emit_cmove): Don't check TARGET_E500.
3253         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3254         sordered): Don't check TARGET_E500.
3255
3256 2007-01-01  Eric Christopher  <echristo@apple.com>
3257
3258         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3259         HARD_REGISTER_NUM_P.
3260
3261 2007-01-01  Roger Sayle  <roger@eyesopen.com>
3262
3263         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3264         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3265         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3266         eq/ne Y".
3267
3268 2007-01-01  Mike Stump  <mrs@apple.com>
3269
3270         * configure.ac: Remove support for building with Apple's gcc-3.1.
3271
3272 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
3273
3274         PR middle-end/30311
3275         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3276         REG is a hard register.  Do not modify REG before calling
3277         subreg_nregs.
3278         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3279         multiple of the size of YMODE for certain lowpart cases.
3280
3281 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3282
3283         PR middle-end/30253
3284         * gimplify (voidify_wrapper_expr): Update for
3285         GIMPLE_MODIFY_STMT.
3286
3287 2007-01-01  Andreas Schwab  <schwab@suse.de>
3288
3289         PR target/29166
3290         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3291         save of BR0 in extra_spill_size instead of spill_size.
3292         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3293         area.
3294         (ia64_expand_epilogue): Restore BR0 from its new location.
3295
3296 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3297
3298         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3299         Use a temporary variable if the left hand side is not a gimple
3300         register.
3301
3302 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
3303
3304         * gimplify.c (gimplify_return_expr): Make the temporary variable
3305         for the return expression, a gimple register variable.
3306
3307 2007-01-01  Jan Hubicka  <jh@suse.cz>
3308
3309         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3310         INSN_LIST.
3311
3312 2007-01-01  Mike Stump  <mrs@apple.com>
3313
3314         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3315         .literal16.
3316         * config/darwin.c (machopic_select_rtx_section): Use
3317         HAVE_GAS_LITERAL16.
3318         (darwin_mergeable_constant_section): Likewise.
3319         * configure: Regenerate.
3320         * config.in: Regenerate.
3321
3322 2007-01-01  Jan Hubicka  <jh@suse.cz>
3323             Andrew Pinski  <pinskia@gmail.com>
3324
3325         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3326         before starting IPA passes.