gcc/ChangeLog

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