OSDN Git Service

* config/darwin.h (LINK_COMMAND_SPEC): Add .cxx/.cp for dsymutil
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
index 52da89f..eb66eff 100644 (file)
+2007-05-24  Ollie Wild  <aaw@google.com>
+
+       * doc/cpp.texi (Common Predefined Macros): Add __COUNTER__
+       description.
+
+2007-05-24  Richard Sandiford  <rsandifo@nildram.co.uk>
+
+       * postreload-gcse.c (reg_changed_after_insn_p): New function.
+       (oprs_unchanged_p): Use it to check all registers in a REG.
+       (record_opr_changes): Look for clobbers in CALL_INSN_FUNCTION_USAGE.
+       (reg_set_between_after_reload_p): Delete.
+       (reg_used_between_after_reload_p): Likewise.
+       (reg_set_or_used_since_bb_start): Likewise.
+       (eliminate_partially_redundant_load): Use reg_changed_after_insn_p
+       and reg_used_between_p instead of reg_set_or_used_since_bb_start.
+       Use reg_set_between_p instead of reg_set_between_after_reload_p.
+       * rtlanal.c (reg_set_p): Check whether REG overlaps
+       regs_invalidated_by_call, rather than just checking the
+       membership of REGNO (REG).
+
+2007-05-24  Zdenek Dvorak  <dvorakz@suse.cz>
+
+       * doc/passes.texi: Document predictive commoning.
+       * doc/invoke.texi (-fpredictive-commoning): Document.
+       * opts.c (decode_options): Enable flag_predictive_commoning on -O3.
+       * tree-ssa-loop-im.c (get_lsm_tmp_name): Export.  Allow
+       adding indices to the generated name.
+       (schedule_sm): Pass 0 to get_lsm_tmp_name.
+       * tree-ssa-loop-niter.c (stmt_dominates_stmt_p): Export.
+       * tree-pretty-print.c (op_symbol_1): Renamed to ...
+       (op_symbol_code): ... and exported.
+       (dump_omp_clause, op_symbol): Use op_symbol_code
+       instead of op_symbol_1.
+       * tree-pass.h (pass_predcom): Declare.
+       * timevar.def (TV_PREDCOM): New timevar.
+       * tree-ssa-loop.c (run_tree_predictive_commoning,
+       gate_tree_predictive_commoning, pass_predcom): New.
+       * tree-data-ref.c (find_data_references_in_loop): Find the
+       references in dominance order.
+       (canonicalize_base_object_address): Ensure that the result has
+       pointer type.
+       (dr_analyze_innermost): Export.
+       (create_data_ref): Code to fail for references with invariant
+       address moved ...
+       (find_data_references_in_stmt): ... here.
+       * tree-data-ref.h (dr_analyze_innermost): Declare.
+       * tree-affine.c: Include tree-gimple.h and hashtab.h.
+       (aff_combination_find_elt, name_expansion_hash,
+       name_expansion_eq, tree_to_aff_combination_expand,
+       double_int_constant_multiple_p, aff_combination_constant_multiple_p):
+       New functions.
+       * tree-affine.h (aff_combination_constant_multiple_p,
+       tree_to_aff_combination_expand): Declare.
+       * tree-predcom.c: New file.
+       * common.opt (fpredictive-commoning): New option.
+       * tree-flow.h (op_symbol_code, tree_predictive_commoning,
+       stmt_dominates_stmt_p, get_lsm_tmp_name): Declare.
+       * Makefile.in (tree-predcom.o): Add.
+       (tree-affine.o): Add TREE_GIMPLE_H dependency.
+       * passes.c (init_optimization_passes):  Add dceloop after
+       copy propagation in loop optimizer.  Add predictive commoning
+       to loop optimizer passes.
+
+2007-05-24  H.J. Lu  <hongjiu.lu@intel.com>
+
+       * target-def.h (TARGET_MANGLE_DECL_ASSEMBLER_NAME): Correct
+       default hook.
+
+2007-05-24  Jan Hubicka  <jh@suse.cz>
+
+       * gengenrtl.c (gendecl, gendef): Output the gens annotated for
+       statistics.
+       (genheader): Include statistics.h.
+
+       * doc/invoke.texi (-fdump-unnumbered): Update docs when line number
+       notes are gone.
+       * print-rtl.c (flag_dump_unnumbered): Update comments.
+       (print_rtl): Fix my previous change.
+       * emit-rtl.c (emit_note_before, emit_note_after): Clear out note
+       specific data.
+
+2007-05-24  Zdenek Dvorak  <dvorakz@suse.cz>
+
+       PR middle-end/32018
+       * tree-ssa-threadupdate.c (thread_through_loop_header): Use
+       set_loop_copy.
+       (thread_through_all_blocks): Call initialize_original_copy_tables
+       and free_original_copy_tables.
+       * cfgloopmanip.c (duplicate_loop, duplicate_loop_to_header_edge):
+       Use set_loop_copy.
+       * tree-cfg.c (tree_duplicate_sese_region): Ditto.
+       * cfghooks.c (duplicate_block): Use get_loop_copy.
+       * cfg.c: Include cfgloop.h.
+       (loop_copy): New hash table.
+       (initialize_original_copy_tables): Initialize loop_copy table.
+       (free_original_copy_tables): Free loop_copy table.
+       (copy_original_table_clear, copy_original_table_set,
+       set_loop_copy, get_loop_copy): New functions.
+       (set_bb_original, set_bb_copy): Use copy_original_table_set.
+       * cfgloop.h (struct loop): Remove copy field.
+       * Makefile.in (cfg.o): Add CFGLOOP_H dependency.
+       * basic-block.h (set_loop_copy, get_loop_copy): Declare.
+
+2007-05-24  H.J. Lu  <hongjiu.lu@intel.com>
+
+       * config/i386/i386.c (ix86_handle_option): Handle SSE4.1 for
+       -msse/-msse2/-msse3.
+
+2007-05-24  H.J. Lu  <hongjiu.lu@intel.com>
+
+       * config/i386/i386.c (ix86_init_mmx_sse_builtins): Mark
+       __builtin_ia32_vec_set_v2di with MASK_64BIT.
+
+2007-05-24  Danny Smith  <dannysmith@users.sourceforge.net>
+
+       PR target/27067
+       * doc/tm.texi (TARGET_MANGLE_DECL_ASSEMBLER_NAME): Document.
+       * targhooks.h (default_mangle_decl_assembler_name): Declare
+       default hook.
+       * targhooks.c (default_mangle_decl_assembler_name): Define
+       default hook.
+       * target-def.h (TARGET_MANGLE_DECL_ASSEMBLER_NAME) New. Set to
+       default hook.
+       * target.h (struct gcc_target): Add mangle_decl_assembler_name field.
+       * langhooks.c (lhd_set_decl_assembler_name): Call
+       targetm.mangle_decl_assembler_name for names with global scope.
+
+       * config/i386/cygming.h (TARGET_MANGLE_DECL_ASSEMBLER_NAME) Override
+       default.
+       (ASM_OUTPUT_DEF_FROM_DECLS): Simplify to use DECL_ASSEMBLER_NAME.
+       * config/i386/i386-protos.h (i386_pe_mangle_decl_assembler_name):
+       Declare.
+       * config/i386/winnt.c (i386_pe_maybe_mangle_decl_assembler_name):
+       New. Factored out of i386_pe_encode_section_info.
+       (gen_stdcall_or_fastcall_suffix): Get name identifier as argument.
+       Move check for prior decoration of stdcall
+       symbols to i386_pe_encode_section_info.
+       (i386_pe_encode_section_info): Adjust call to
+       gen_stdcall_or_fastcall_suffix.  Use
+       i386_pe_maybe_mangle_decl_assembler_name, if needed.
+       (i386_pe_mangle_decl_assembler_name): New. Wrap
+       i386_pe_maybe_mangle_decl_assembler_name.
+
+2007-05-16  Rafael Avila de Espindola  <espindola@google.com>
+
+       * c-common.c (c_common_signed_or_unsigned_type): Delay the check for
+       INTEGRAL_TYPE_P and TYPE_UNSIGNED.
+       * langhooks.c (get_signed_or_unsigned_type): Don't check for
+       INTEGRAL_TYPE_P or TYPE_UNSIGNED.
+       (lhd_signed_or_unsigned_type): Check for INTEGRAL_TYPE_P and
+       TYPE_UNSIGNED.
+
+2007-05-23  Sandra Loosemore  <sandra@codesourcery.com>
+           Nigel Stephens  <nigel@mips.com>
+           Richard Sandiford  <richard@codesourcery.com>
+
+       Fix up MIPS16 hard float and add support for complex.
+
+       * config/mips/mips.h (TARGET_HARD_FLOAT_ABI): New.
+       (TARGET_SOFT_FLOAT_ABI): New.
+       (TARGET_CPU_CPP_BUILTINS): Define __mips_hard_float and
+       __mips_soft_float to reflect the ABI in use, not whether the
+       FPU is directly accessible (e.g., in MIPS16 mode).
+       (UNITS_PER_HWFPVALUE): Use TARGET_SOFT_FLOAT_ABI.
+       (UNITS_PER_FPVALUE): Likewise.
+
+       * config/mips/mips.c (mips_expand_call): Remove redundant
+       TARGET_MIPS16 check.
+       (mips_arg_regno): New.
+       (function_arg_advance): When setting bits in cum->fp_code for
+       MIPS16, don't subtract 1 from cum->arg_number, since it is now
+       zero-based.
+       (function_arg): Use mips_arg_regno.
+       (mips_return_mode_in_fpr_p): New.
+       (mips16_call_stub_mode_suffix): New.
+       (mips16_cfun_returns_in_fpr_p): New.
+       (mips_save_reg_p): Use mips16_cfun_returns_in_fpr_p.
+       (mips_output_function_prologue): Test mips16_hard_float, not
+       !TARGET_SOFT_FLOAT, to decide when a function stub is required.
+       (mips_expand_epilogue): Call MIPS16 helper routines to copy
+       return value into a floating-point register.
+       (mips_can_use_return_insn): Use mips16_cfun_returns_in_fpr_p.
+       (mips_function_value): Rewrite to use mips_return_mode_in_fpr_p.
+       (mips16_fp_args): Handle MIPS32r2 ISA which supports
+       TARGET_FLOAT64, and use mfhc1/mthc1 to copy the most significant
+       word of double arguments from or to the high bits of 64-bit
+       floating point registers.
+       (build_mips16_function_stub): Fill in DECL_RESULT for stubdecl.
+       (mips16_fpret_double): New helper function.
+       (build_mips16_call_stub): Use mips16_return_mode_in_fpr_p.  Add
+       support for complex modes.  Fill in DECL_RESULT for stubdecl.
+       (mips_init_libfuncs): Remove redundant TARGET_MIPS16 check.
+
+       * config/mips/mips16.S 
+       (RET, ARG1, ARG2): New.
+       (MERGE_GPRf, MERGE_GPRt): New.
+       (DELAYt, DELAYf): New.
+       (MOVE_SF_BYTE0, MOVE_SI_BYTE0): New.
+       (MOVE_SF_BYTE4, MOVE_SF_BYTE8): New.
+       (MOVE_DF_BYTE0, MOVE_DF_BYTE8): New.
+       (MOVE_SF_RET, MOVE_SC_RET, MOVE_DF_RET, MOVE_DC_RET, MOVE_SI_RET): New.
+       (SFOP): Renamed to...
+       (OPSF3): This, and macro-ified.  Updated all uses.
+       (SFOP2): Renamed to...
+       (OPSF2): This, and macro-ified.  Updated all uses.
+       (SFCMP): Renamed to...
+       (CMPSF): This, and macro-ified.  Updated all uses.
+       (SFREVCMP): Renamed to...
+       (REVCMPSF): This, and macro-ified.  Updated all uses.
+       (__mips16_floatsisf, __mips16_fix_truncsfsi): Macro-ified.
+       (LDDBL1, LDDBL2, RETDBL): Deleted.
+       (DFOP): Renamed to...
+       (OPDF3): This, and macro-ified.  Updated all uses.
+       (DFOP2): Renamed to...
+       (OPDF2): This, and macro-ified.  Updated all uses.
+       (__mips16_extendsfdf2, __mips16_truncdfsf2): Macro-ified.
+       (DFCMP): Renamed to...
+       (CMPDF): This, and macro-ified.  Updated all uses.
+       (DFREVCMP): Renamed to...
+       (REVCMPDF): This, and macro-ified.  Updated all uses.
+       (__mips16_floatsidf, __mips16_fix_truncdfsi): Macro-ified.
+       (RET_FUNCTION): New.
+       (__mips16_ret_sf, __mips16_ret_df): Macro-ified.
+       (__mips16_ret_sc, __mips16_ret_dc): New.
+       (STUB_ARGS_0, STUB_ARGS_1, STUB_ARGS_5, STUB_ARGS_9, STUB_ARGS_2,
+       STUB_ARGS_6, STUB_ARGS_10): New.
+       (CALL_STUB_NO_RET): New.
+       (__mips16_call_stub_1): Macro-ified.
+       (__mips16_call_stub_5): Macro-ified.
+       (__mips16_call_stub_2): Macro-ified.
+       (__mips16_call_stub_6): Macro-ified.
+       (__mips16_call_stub_9): Macro-ified.
+       (__mips16_call_stub_10): Macro-ified.
+       (CALL_STUB_RET): New.
+       (__mips16_call_stub_sf_0): Macro-ified.
+       (__mips16_call_stub_sf_1): Macro-ified.
+       (__mips16_call_stub_sf_5): Macro-ified.
+       (__mips16_call_stub_sf_2): Macro-ified.
+       (__mips16_call_stub_sf_6): Macro-ified.
+       (__mips16_call_stub_sf_9): Macro-ified.
+       (__mips16_call_stub_sf_10): Macro-ified.
+       (__mips16_call_stub_df_0): Macro-ified.
+       (__mips16_call_stub_df_1): Macro-ified.
+       (__mips16_call_stub_df_5): Macro-ified.
+       (__mips16_call_stub_df_2): Macro-ified.
+       (__mips16_call_stub_df_6): Macro-ified.
+       (__mips16_call_stub_df_9): Macro-ified.
+       (__mips16_call_stub_df_10): Macro-ified.
+       (__mips16_call_stub_sc_0): New.
+       (__mips16_call_stub_sc_1): New.
+       (__mips16_call_stub_sc_5): New.
+       (__mips16_call_stub_sc_2): New.
+       (__mips16_call_stub_sc_6): New.
+       (__mips16_call_stub_sc_9): New.
+       (__mips16_call_stub_sc_10): New.
+       (__mips16_call_stub_dc_0): New.
+       (__mips16_call_stub_dc_1): New.
+       (__mips16_call_stub_dc_5): New.
+       (__mips16_call_stub_dc_2): New.
+       (__mips16_call_stub_dc_6): New.
+       (__mips16_call_stub_dc_9): New.
+       (__mips16_call_stub_dc_10): New.
+       
+       * config/mips/t-elf (LIB1ASMFUNCS): Add MIPS16 floating-point stubs.
+       * config/mips/t-isa3264 (LIB1ASMFUNCS): Likewise.
+       * config/mips/t-r2900 (LIB1ASMFUNCS): Likewise.
+
+2007-05-23  Ian Lance Taylor  <iant@google.com>
+
+       * doc/invoke.texi (Invoking GCC): Document that the order of the
+       -l option matters.
+
 2007-05-23  Chen Liqin  <liqin@sunnorth.com.cn>
 
        PR target/30987
-       * config/score/misc.md (bitclr_c, bitset_c, bittgl_c): remove.
-       * config/score/predicate.md (const_pow2, const_npow2): remove.
-       * config/score/score.h (ASM_OUTPUT_EXTERNAL): add ASM_OUTPUT_EXTERNAL undef.
+       * config/score/misc.md (bitclr_c, bitset_c, bittgl_c): Remove.
+       * config/score/predicate.md (const_pow2, const_npow2): Remove.
+       * config/score/score.h (ASM_OUTPUT_EXTERNAL): Add ASM_OUTPUT_EXTERNAL
+       undef.
+
        PR target/30474
-       * config/score/score.c (score_print_operand): makes sure that only lower 
-        bits are used.
+       * config/score/score.c (score_print_operand): Make sure that only
+       lower bits are used.
        
 2007-05-22  Ian Lance Taylor  <iant@google.com>
 
        Add loopno argument.  Use evolution_function_is_invariant_rec_p
        instead of evolution_function_is_constant_p.
        Update calls to evolution_function_is_affine_multivariate_p.
-       * tree-chrec.h  (evolution_function_is_affine_multivariate_p):
+       * tree-chrec.h (evolution_function_is_affine_multivariate_p):
        Add loopno argument.
        * tree-scalar-evolution.c (gather_chrec_stats): Call
-       evolution_function_is_affine_multivariate_p with a  loop
+       evolution_function_is_affine_multivariate_p with a loop
        number of 0.
        * tree-data-ref.c (analyze_miv_subscript): Likewise.
        (analyze_overlapping_iterations): Likewise.
-       (access_functions_are_affine_or_constant_p):
-       Likewise.
+       (access_functions_are_affine_or_constant_p): Likewise.
        (build_classic_dist_vector_1): If the access functions
        are equal, don't do anything.
 
 2007-05-21  Paolo Bonzini  <bonzini@gnu.org>
-            Paolo Carlini  <pcarlini@suse.de>
+           Paolo Carlini  <pcarlini@suse.de>
            Uros Bizjak  <ubizjak@gmail.com>
 
        * c-cppbuiltin.c (c_cpp_builtins): Define
 
 2007-05-21  Mike Stump  <mrs@apple.com>
 
-       * config/darwin.h (LINK_COMMAND_SPEC): Add .cxx for dsymutil
+       * config/darwin.h (LINK_COMMAND_SPEC): Add .cxx/.cp for dsymutil
        handling as well.
        * config/darwin9.h (LINK_COMMAND_SPEC): Likewise.
 
        NOTE_INSN_BASIC_BLOCK_P when possible.
        * ifcvt.c (dead_or_predicable): Remove call of squeeze_notes.
        * dwarf2out.c (gen_label_die): Use NOTE_KIND instead of
-        NOTE_LINE_NUMBER; use NOTE_INSN_BASIC_BLOCK_P when possible.
+       NOTE_LINE_NUMBER; use NOTE_INSN_BASIC_BLOCK_P when possible.
        (dwarf2out_var_location): Likewise.
        * cfgbuild.c (make_edges): Likewise.
        (find_basic_blocks_1): Likewise.