compiler: Evaluate array range expression once before iteration.

Fixes issue 34.

From-SVN: r219045

Fix for PR ipa/63851 and ipa/63852.

PR ipa/63851
PR ipa/63852
* ipa-icf.c (sem_function::merge): Ignore merge operation
for a thunk created from static chain.
* ipa-icf-gimple.c (func_checker::compatible_types_p): Verify that
types have same restrict flag.

From-SVN: r219042

Daily bump.

From-SVN: r219041

re PR target/55023 (hppa: wrong code generated with tail call optimisation)

PR target/55023
* dse.c (scan_insn): Treat sibling call as though it does a wild read.
* testsuite/gcc.dg/pr55023.c: New file.

From-SVN: r219037

* lib/ubsan-dg.exp: Add library path for libstdc++.

From-SVN: r219035

trans-intrinsic.c (gfc_conv_intrinsic_caf_get, [...]): Fix vector handling.

2014-12-22  Tobias Burnus  <burnus@net-b.de>

        * trans-intrinsic.c (gfc_conv_intrinsic_caf_get, conv_caf_send):
        Fix vector handling.

From-SVN: r219034

basic_string.h: Fix _S_local_capacity for large sizeof(_CharT).

2014-12-22  Paolo Carlini  <paolo.carlini@oracle.com>

* include/bits/basic_string.h: Fix _S_local_capacity for
large sizeof(_CharT).

From-SVN: r219031

re PR target/58314 (SH4 error: 'asm' operand requires impossible reload)

gcc/testsuite/
PR target/58314
* gcc.target/sh/torture/pr58314-2.c: New.
* gcc.target/sh/torture/pr58314.c: Don't set -Os option.

From-SVN: r219030

re PR fortran/63363 (No diagnostic for passing function as actual argument to KIND)

2014-12-22  Janus Weil  <janus@gcc.gnu.org>

PR fortran/63363
* check.c (gfc_check_kind): Reject polymorphic and non-data arguments.

2014-12-22  Janus Weil  <janus@gcc.gnu.org>

PR fortran/63363
* gfortran.dg/kind_1.f90: New.

From-SVN: r219027

re PR libstdc++/37522 (Incorrect vswprintf prototype breaks __to_xstring)

PR libstdc++/37522
* include/bits/basic_string.h (stod, stof, stoi, stol, stold, stoll,
stoul, stoull, to_string): Only use _GLIBCXX_HAVE_BROKEN_VSWPRINTF
to guard definition of to_wstring.
* testsuite/21_strings/basic_string/numeric_conversions/char/dr1261.cc:
Do not use dg-require-string-conversions.
* testsuite/21_strings/basic_string/numeric_conversions/char/stod.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stof.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stoi.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stol.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stold.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stoll.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stoul.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/
stoull.cc: Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/
to_string.cc: Likewise.

From-SVN: r219019

re PR target/52933 (SH Target: Use div0s for integer sign comparisons)

gcc/testsuite/
PR target/52933
* gcc.target/sh/sh/pr52933-3.c: New.

From-SVN: r219017

re PR libstdc++/60271 ([DR2369] [C++1y] std::max(initializer_list<T>) cannot use std::max_element)

2014-12-22  Ville Voutilainen  <ville.voutilainen@gmail.com>

PR libstdc++/60271
C++14 constexpr min, max, minmax, min_element, max_element
and minmax_element. Also constexpr for 20.9.5-20.9.9,
aka various library functors.
* include/bits/c++config: Add _GLIBCXX14_CONSTEXPR.
* include/bits/algorithmfwd.h (min, max, minmax, min_element,
max_element): Use it.
* include/bits/predefined_ops.h (_Iter_less_iter, __iter_less_iter,
_Iter_comp_iter, __iter_comp_iter): Likewise.
* include/bits/stl_algo.h (minmax, __minmax_element, minmax_element,
min, max, __min_element, min_element, __max_element, max_element)
Likewise.
* include/bits/stl_algobase.h (min, max): Likewise.
* include/bits/stl_function.h (plus, minus, multiplies, divides,
modulus, negate, equal_to, not_equal_to, greater, less, greater_equal,
less_equal, logical_and, logical_or, logical_not, bit_and, bit_or,
bit_xor, bit_not, unary_negate, not1, binary_negate, not2): Likewise.
* testsuite/20_util/function_objects/constexpr.cc: New.
* testsuite/25_algorithms/max/constexpr.cc: Likewise.
* testsuite/25_algorithms/max_element/constexpr.cc: Likewise.
* testsuite/25_algorithms/min/constexpr.cc: Likewise.
* testsuite/25_algorithms/min_element/constexpr.cc: Likewise.
* testsuite/25_algorithms/minmax/constexpr.cc: Likewise.
* testsuite/25_algorithms/minmax_element/constexpr.cc: Likewise.
* testsuite/ext/profile/mutex_extensions_neg.cc: Adjust dg-error.

From-SVN: r219015

status_cxx2011.xml: Update.

* doc/xml/manual/status_cxx2011.xml: Update.
* doc/html/manual/status.html: Regenerate.

From-SVN: r219014

re PR libstdc++/54354 (TODO extended iomanip manipulators std::get_time and std::put_time (C++11, section 27.7.5))

2014-12-22  Rüdiger Sonderfeld  <ruediger@c-plusplus.de>

PR libstdc++/54354
* include/std/iomanip (_Get_time): New struct.
(get_time): New manipulator.
(operator<<): New overloaded function.
* testsuite/27_io/manipulators/extended/get_time/char/1.cc: New.
* testsuite/27_io/manipulators/extended/get_time/char/2.cc: New.
* testsuite/27_io/manipulators/extended/get_time/wchar_t/1.cc: New.
* testsuite/27_io/manipulators/extended/get_time/wchar_t/2.cc: New.

From-SVN: r219013

re PR libstdc++/60396 (Missing time_get<>::get() functions)

2014-12-22  Rüdiger Sonderfeld  <ruediger@c-plusplus.de>
    Jonathan Wakely  <jwakely@redhat.com>

PR libstdc++/60396
* config/abi/pre/gnu.ver: Add new exports.
* include/bits/locale_facets_nonio.h (time_get::get): New overloads.
(time_get::do_get):  New virtual function.
* include/bits/locale_facets_nonio.tcc (time_get::get): Define.
(time_get::do_get): Define.
* src/c++11/string-inst.cc (time_get::get, time_get::do_get): Add
C++11-only instantiations.
* testsuite/22_locale/time_get/get/char/1.cc: New.
* testsuite/22_locale/time_get/get/char/2.cc: New.
* testsuite/22_locale/time_get/get/wchar_t/1.cc: New.
* testsuite/22_locale/time_get/get/wchar_t/2.cc: New.

Co-Authored-By: Jonathan Wakely <jwakely@redhat.com>
From-SVN: r219012

* gnat.dg/nan_max.adb: New test.

From-SVN: r219011

trans.c (Attribute_to_gnu): If the type is a floating-point type...

* gcc-interface/trans.c (Attribute_to_gnu) <Attr_{Min,Max}>: If the
type is a floating-point type, implement the semantics of the C99
f{min,max} routines with regard to NaNs.
(gnat_to_gnu): Call builtin_decl_implicit.
* gcc-interface/utils2.c (compare_arrays): Adjust comments.

From-SVN: r219010

decl.c (gnat_to_gnu_field): Post the error message for parent overlapping on the position instead of on the...

* gcc-interface/decl.c (gnat_to_gnu_field): Post the error message
for parent overlapping on the position instead of on the first bit.
For a field that needs strict alignment, issue the error for the
position first and, for the size, issue an error if it is too large
only for the atomic and aliased cases.  Issue a specific error if
the size is not a multiple of a byte in the volatile and the stric
alignment cases.

From-SVN: r219009

re PR rtl-optimization/62151 (wrong code at -O2 and -O3 on x86_64-linux-gnu)

PR rtl-optimization/62151
* combine.c (try_combine): New local variables local_elim_i1
and local_elim_i0.  Set elim_i1 and elim_i0 using the local
version variables.  Distribute notes from i0notes or i1notes
using the local variables.

gcc/testsuite/ChangeLog
PR rtl-optimization/62151
* gcc.c-torture/execute/pr62151.c: New test.

From-SVN: r219008

decl.c (check_ok_for_atomic): Rename into...

* gcc-interface/decl.c (check_ok_for_atomic): Rename into...
(check_ok_for_atomic_type): ...this.  When checking the mode, also
check that the type is sufficient aligned.  Remove useless code and
tidy up implementation.
(gnat_to_gnu_entity): Adjust to above renaming.
(gnat_to_gnu_component_type): Likewise.
(gnat_to_gnu_field): Likewise.

From-SVN: r219007

symbol_summary is used for inline_summary.

* lto-partition.c (add_symbol_to_partition_1): New inline_summaries
is used.
(undo_partition): Likewise.
(lto_balanced_map): Likewise.
* cgraphunit.c (symbol_table::process_new_functions): New inline_summaries
is used.
* ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
(devirtualization_time_bonus): Likewise.
(estimate_local_effects): Likewise.
(ipcp_propagate_stage): Likewise.
* ipa-inline-analysis.c (evaluate_conditions_for_known_args): Likewise.
(evaluate_properties_for_edge): Likewise.
(inline_summary_alloc): Likewise.
(reset_inline_summary): New inline_summary argument is introduced.
(inline_summary_t::remove): New function.
(inline_summary_t::duplicate): Likewise.
(dump_inline_edge_summary): New inline_summaries is used.
(dump_inline_summary): Likewise.
(estimate_function_body_sizes): Likewise.
(compute_inline_parameters): Likewise.
(estimate_edge_devirt_benefit): Likewise.
(estimate_node_size_and_time): Likewise.
(inline_update_callee_summaries): Likewise.
(inline_merge_summary): Likewise.
(inline_update_overall_summary): Likewise.
(simple_edge_hints): Likewise.
(do_estimate_edge_time): Likewise.
(estimate_time_after_inlining): Likewise.
(estimate_size_after_inlining): Likewise.
(do_estimate_growth): Likewise.
(growth_likely_positive): Likewise.
(inline_generate_summary): Likewise.
(inline_read_section): Likewise.
(inline_read_summary): Likewise.
(inline_write_summary): Likewise.
(inline_free_summary): Likewise.
* ipa-inline-transform.c (clone_inlined_nodes): Likewise.
(inline_call): Likewise.
* ipa-inline.c (caller_growth_limits): Likewise.
(can_inline_edge_p): Likewise.
(want_early_inline_function_p): Likewise.
(compute_uninlined_call_time): Likewise.
(compute_inlined_call_time): Likewise.
(big_speedup_p): Likewise.
(want_inline_small_function_p): Likewise.
(edge_badness): Likewise.
(update_caller_keys): Likewise.
(update_callee_keys): Likewise.
(recursive_inlining): Likewise.
(inline_small_functions): Likewise.
(inline_to_all_callers): Likewise.
(dump_overall_stats): Likewise.
(early_inline_small_functions): Likewise.
* ipa-inline.h: New class inline_summary_t replaces
vec<inline_summary_t>.
* ipa-split.c (execute_split_functions): New inline_summaries is used.
* ipa.c (walk_polymorphic_call_targets): Likewise.
* tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.

From-SVN: r219006

ipa-prop uses symbol_summary class.

* lto-partition.c: Include of symbol-summary.h is added.
* lto-symtab.c: Likewise.
* lto.c: Likewise.
* auto-profile.c: Include of symbol-summary.h is added.
* cgraph.c: Likewise.
* cgraphbuild.c: Likewise.
* cgraphclones.c: Likewise.
* cgraphunit.c: Likewise.
* ipa-cp.c: Likewise.
* ipa-devirt.c: Likewise.
* ipa-icf.c: Likewise.
* ipa-inline-analysis.c (evaluate_properties_for_edge): New
ipa_node_params_sum data structure is used.
(inline_node_duplication_hook): Likewise.
(estimate_function_body_sizes): Likewise.
(remap_edge_change_prob): Likewise.
(inline_merge_summary): Likewise.
* ipa-inline-transform.c: Include of symbol-summary.h is added.
* ipa-inline.c (early_inliner): New ipa_node_params_sum data structure
is used.
* ipa-polymorphic-call.c: Include of symbol-summary.h is added.
* ipa-profile.c: Include of symbol-summary.h is added.
* ipa-prop.c (ipa_propagate_indirect_call_infos): New ipa_node_params_sum
data structure is used.
(ipa_node_params::~ipa_node_params): New function.
(ipa_free_all_node_params): Destruction is simplified.
(ipa_node_removal_hook): Removed.
(ipa_add_new_function): Renamed from ipa_node_duplication_hook.
(ipa_node_params_t::duplicate): New function.
(ipa_register_cgraph_hooks): Few hooks are removed.
(ipa_unregister_cgraph_hooks): Likewise.
(ipa_prop_write_jump_functions): New ipa_node_params_sum is used.
* ipa-prop.h (struct ipa_node_params): Destructor introduced for
the structure.
(ipa_check_create_node_params): Vector for ipa_node_params is replaced
with function_summary.
* ipa-split.c: Include of symbol-summary.h is added.
* ipa-utils.c: Include of symbol-summary.h is added.
* ipa.c: Include of symbol-summary.h is added.
* omp-low.c: Include of symbol-summary.h is added.
* tree-inline.c: Include of symbol-summary.h is added.
* tree-sra.c: Include of symbol-summary.h is added.
* tree-ssa-pre.c: Include of symbol-summary.h is added.

From-SVN: r219005

New symbol_summary class introduced.

* cgraph.h (symbol_table::allocate_cgraph_symbol): Summary UID
is filled up.
* symbol-summary.h: New file.
* gengtype.c (open_base_files): Add symbol-summary.h.
* toplev.c (general_init): Call constructor of symbol_table.

From-SVN: r219004

Add myself to RTEMS Ports maintainers.

ChangeLog
* MAINTAINERS (RTEMS Ports): Add myself.

From-SVN: r219003

Daily bump.

From-SVN: r219002

re PR target/55212 ([SH] Switch to LRA)

gcc/
PR target/55212
* config/sh/sh.md (*addsi3_compact): Add parentheses around &&
condition.  Add comments.

From-SVN: r218999

re PR target/17280 (-mrelax and -O cause .uses directives in assembler to refer to non-existent local labels)

gcc/testsuite/
PR target/17280
* gcc.target/sh/torture/pr17280.c: New.

From-SVN: r218998

re PR libstdc++/64367 (g++-v5/stdexcept:52:28: error: invalid use of non-static data member '_M_p')

PR libstdc++/64367
* include/std/stdexcept: Don't use non-static member in sizeof.

From-SVN: r218997

re PR c++/64359 (alias_template_specialization_p & template_args_equal show up high in profile)

PR c++/64359
* pt.c (iterative_hash_template_arg): Hash alias specializations
differently from their TYPE_CANONICAL.
(alias_template_specialization_p): Optimize.
(template_args_equal): Optimize alias handling.
(dependent_alias_template_spec_p): Only check innermost args.

From-SVN: r218995

Daily bump.

From-SVN: r218994

vector (std::__debug::vector<>::clear()): Do not reset guaranteed capacity.

2014-12-20  François Dumont  <fdumont@gcc.gnu.org>

* include/debug/vector (std::__debug::vector<>::clear()): Do not reset
guaranteed capacity.
* testsuite/23_containers/vector/bool/allocator/swap.cc: Move
propagating_allocator equality and inequality operators to __gnu_test
namespace.

From-SVN: r218991

mutex_extensions_neg.cc: Adjust dg-error.

2014-12-20  François Dumont  <fdumont@gcc.gnu.org>

* testsuite/ext/profile/mutex_extensions_neg.cc: Adjust dg-error.

From-SVN: r218990

re PR target/64358 (Wrong code for __int128 operations in powerpc64le)

PR target/64358
* config/rs6000/rs6000.c (rs6000_split_logical_inner): Swap the
input operands if only the second is inverted.
* config/rs6000/rs6000.md (*boolc<mode>3_internal1 for BOOL_128):
Swap BOOL_REGS_OP1 and BOOL_REGS_OP2.  Correct arguments to
rs6000_split_logical.
(*boolc<mode>3_internal2 for TI2): Swap operands[1] and operands[2].

From-SVN: r218989

c++config: inline __cxx11 namespace within namespace __cxx1998.

2014-12-20  François Dumont  <fdumont@gcc.gnu.org>

* include/bits/c++config: inline __cxx11 namespace within namespace
__cxx1998.

From-SVN: r218988

* g++.dg/cpp1y/lambda-generic-variadic2.C: Fix.

From-SVN: r218987

c-cppbuiltin.c (__cpp_sized_deallocation): Uncomment and move macro.

c-family/

2014-12-20  Edward Smith-Rowland  <3dw4rd@verizon.net>

* c-cppbuiltin.c (__cpp_sized_deallocation): Uncomment and move macro.
Control macro with flag_sized_deallocation.

testsuite/

2014-12-20  Edward Smith-Rowland  <3dw4rd@verizon.net>

* g++.dg/cpp1y/feat-cxx98-neg.C: Enable __cpp_sized_deallocation test.
* g++.dg/cpp1y/feat-cxx14.C: Ditto.
* g++.dg/cpp1y/feat-cxx11-neg.C: Ditto and move tests for consistent
order.
* g++.dg/cpp1y/feat-sized-dealloc-neg.C: New.
* g++.dg/cpp1y/feat-sized-dealloc.C: New.

From-SVN: r218986

invoke.texi: Document -Wdiscarded-array-qualifiers.

2014-12-20  Martin Uecker <uecker@eecs.berkeley.edu>

* doc/invoke.texi: Document -Wdiscarded-array-qualifiers.
* doc/extend.texi: Document new behavior for pointers to arrays
with qualifiers.

c/
* c-typeck.c: New behavious for pointers to arrays with qualifiers
        (common-pointer-type): For pointers to arrays take qualifiers from
element type.
        (build_conditional_expr): Add warnings for lost qualifiers.
(comp-target-types): Allow pointers to arrays with different qualifiers.
        (convert-for-assignment): Adapt warnings for discarded qualifiers. Add
WARNING_FOR_QUALIFIERS macro and rename WARN_FOR_QUALIFIERS
        to PEDWARN_FOR_QUALIFIERS.

c-family/
* c.opt (Wdiscarded-array-qualifiers): New option.

testsuite/
* gcc.dg/Wwrite-strings-1.c: Change dg-warning.
* gcc.dg/array-quals-1.c: Use -Wno-discarded-array-qualifiers.
* gcc.dg/array-quals-2.c: Change dg-options, dg-warning.
* gcc.dg/pointer-array-atomic.c: New test.
* gcc.dg/pointer-array-quals-1.c: New test.
* gcc.dg/pointer-array-quals-2.c: New test (-pedantic-errors).
* gcc.dg/qual-component-1.c: Change dg-options, dg-warnings.

From-SVN: r218985

Fix duplicate symbol errors in Darwin bootstrap.

* src/c++98/locale-inst.cc (__add_grouping, __pad, __int_to_char):
Only compile for old ABI.

From-SVN: r218984

Daily bump.

From-SVN: r218983

compiler: Add temporaries required in cases of interface conversion.

From-SVN: r218979

Fix casting non-"int" to bool.

gcc/jit/ChangeLog:
* jit-playback.c (gcc::jit::playback::context::build_cast): In
case BOOLEAN_TYPE, don't assume that the source expression is
of type "int".

gcc/testsuite/ChangeLog:
* jit.dg/test-expressions.c (make_tests_of_casts): Add tests of
casting between "long" and "bool".
(verify_casts): Verify these new test cases.

From-SVN: r218977

hash-table.h (struct pointer_hash): Fix formating.

* hash-table.h (struct pointer_hash): Fix formating.
(hash_table_higher_prime_index): Declare pure.
(hash_table_mod2, hash_table_mod1, mul_mod): Move inline;
assume that uint64_t always exists.
(hash_table<Descriptor, Allocator, false>): Use gcc_checking_assert.
(hash_table<Descriptor, Allocator, false>::expand ()): Fix formating.
(hash_table<Descriptor, Allocator, false>::clear_slot (value_type **slot)):
Use checking assert.
* hash-table.c: Remove #if 0 code.
(hash_table_higher_prime_index): Use gcc_assert.
(mul_mod, hash-table_mod1, hash_table_mod2): move to hash-table.h

From-SVN: r218976

MIPSR6: mips-img-elf mips-img-linux-gnu triplets and vendor updates

gcc/

* config.gcc: Support mips*-img-linux* and mips*-img-elf*.
* config/mips/mti-linux.h: Support mips32r6 as being the default arch.
* config/mips/t-img-elf: New.
* config/mips/t-img-linux: New.

gcc/testsuite/

* gcc.target/mips/pr37362.c: Skip for mips-img-elf.

From-SVN: r218975

go-gcc.cc (array_constructor_expression): Don't construct arrays of zero-sized values.

* go-gcc.cc (array_constructor_expression): Don't construct arrays
of zero-sized values.

From-SVN: r218974

MIPS32R6 and MIPS64R6 support

gcc/

* config.gcc: Add mipsisa64r6 and mipsisa32r6 cpu support.
* config/mips/constraints.md (ZD): Add r6 restrictions.
* config/mips/gnu-user.h (DRIVER_SELF_SPECS): Add MIPS_ISA_LEVEL_SPEC.
* config/mips/loongson.md
(<u>div<mode>3, <u>mod<mode>3): Move to mips.md.
* config/mips/mips-cpus.def (mips32r6, mips64r6): Define.
* config/mips/mips-modes.def (CCF): New mode.
* config/mips/mips-protos.h
(mips_9bit_offset_address_p): New prototype.
* config/mips/mips-tables.opt: Regenerate.
* config/mips/mips.c (MIPS_JR): Use JALR $, <reg> for R6.
(mips_rtx_cost_data): Add pseudo-processors W32 and W64.
(mips_9bit_offset_address_p): New function.
(mips_rtx_costs): Account for R6 multiply and FMA instructions.
(mips_emit_compare): Implement R6 FPU comparisons.
(mips_expand_conditional_move): Implement R6 selects.
(mips_expand_conditional_trap): Account for removed trap immediate.
(mips_expand_block_move): Disable inline move when LWL/LWR are removed.
(mips_print_float_branch_condition): Update for R6 FPU branches.
(mips_print_operand): Handle CCF mode compares.
(mips_interrupt_extra_call_saved_reg_p): Do not attempt to callee-save
MD_REGS for R6.
(mips_hard_regno_mode_ok_p): Support CCF mode.
(mips_mode_ok_for_mov_fmt_p): Likewise.
(mips_secondary_reload_class): CCFmode can be loaded directly.
(mips_set_fast_mult_zero_zero_p): Account for R6 multiply instructions.
(mips_option_override): Ensure R6 is used with fp64.  Set default
mips_nan modes.  Check for mips_nan support.  Prevent DSP with R6.
(mips_conditional_register_usage): Disable MD_REGS for R6. Disable
FPSW for R6.
(mips_mulsidi3_gen_fn): Support R6 multiply instructions.
* config/mips/mips.h (ISA_MIPS32R6, ISA_MIPS64R6): Define.
(TARGET_CPU_CPP_BUILTINS): Rework for mips32/mips64.
(ISA_HAS_JR): New macro.
(ISA_HAS_HILO): New macro.
(ISA_HAS_R6MUL): Likewise.
(ISA_HAS_R6DMUL): Likewise.
(ISA_HAS_R6DIV): Likewise.
(ISA_HAS_R6DDIV): Likewise.
(ISA_HAS_CCF): Likewise.
(ISA_HAS_SEL): Likewise.
(ISA_HAS_COND_TRAPI): Likewise.
(ISA_HAS_FP_MADDF_MSUBF): Likewise.
(ISA_HAS_LWL_LWR): Likewise.
(ISA_HAS_IEEE_754_LEGACY): Likewise.
(ISA_HAS_IEEE_754_2008): Likewise.
(ISA_HAS_PREFETCH_9BIT): Likewise.
(MIPSR6_9BIT_OFFSET_P): New macro.
(BASE_DRIVER_SELF_SPECS): Use MIPS_ISA_DRIVER_SELF_SPECS.
(DRIVER_SELF_SPECS): Use MIPS_ISA_LEVEL_SPEC.
(MULTILIB_ISA_DEFAULT): Handle mips32r6 and mips64r6.
(MIPS_ISA_LEVEL_SPEC): Likewise.
(MIPS_ISA_SYNCI_SPEC): Likewise.
(ISA_HAS_64BIT_REGS): Likewise.
(ISA_HAS_BRANCHLIKELY): Likewise.
(ISA_HAS_MUL3): Likewise.
(ISA_HAS_DMULT): Likewise.
(ISA_HAS_DDIV): Likewise.
(ISA_HAS_DIV): Likewise.
(ISA_HAS_MULT): Likewise.
(ISA_HAS_FP_CONDMOVE): Likewise.
(ISA_HAS_8CC): Likewise.
(ISA_HAS_FP4): Likewise.
(ISA_HAS_PAIRED_SINGLE): Likewise.
(ISA_HAS_MADD_MSUB): Likewise.
(ISA_HAS_FP_RECIP_RSQRT): Likewise.
* config/mips/mips.md (processor): Add w32 and w64.
(FPCC): New mode iterator.
(reg): Add CCF mode.
(fpcmp): New mode attribute.
(fcond): Add ordered, ltgt and ne codes.
(fcond): Update code attribute.
(sel): New code attribute.
(selinv): Likewise.
(ctrap<mode>4): Update condition.
(*conditional_trap_reg<mode>): New define_insn.
(*conditional_trap<mode>): Update condition.
(mul<mode>3): Expand R6 multiply instructions.
(<su>mulsi3_highpart): Likewise.
(<su>muldi3_highpart): Likewise.
(mul<mode>3_mul3_loongson): Rename...
(mul<mode>3_mul3_hilo): To this.  Add R6 mul instruction.
(<u>mulsidi3_32bit_r6): New expander.
(<u>mulsidi3_32bit): Restrict to pre-r6 multiplies.
(<u>mulsidi3_32bit_r4000): Likewise.
(<u>mulsidi3_64bit): Likewise.
(<su>mulsi3_highpart_internal): Likewise.
(mulsidi3_64bit_r6dmul): New instruction.
(<su>mulsi3_highpart_r6): Likewise.
(<su>muldi3_highpart_r6): Likewise.
(fma<mode>4): Likewise.
(movccf): Likewise.
(*sel<code><GPR:mode>_using_<GPR2:mode>): Likewise.
(*sel<mode>): Likewise.
(<u>div<mode>3): Moved from loongson.md.  Add R6 instructions.
(<u>mod<mode>3): Likewise.
(extvmisalign<mode>): Require ISA_HAS_LWL_LWR.
(extzvmisalign<mode>): Likewise.
(insvmisalign<mode>): Likewise.
(mips_cache): Account for R6 displacement field sizes.
(*branch_fp): Rename...
(*branch_fp_<mode>): To this.  Add CCFmode support.
(*branch_fp_inverted): Rename...
(*branch_fp_inverted_<mode>): To this.  Add CCFmode support.
(s<code>_<mode>): Rename...
(s<code>_<SCALARF:mode>_using_<FPCC:mode>): To this.  Add FCCmode
condition support.
(s<code>_<mode> swapped): Rename...
(s<code>_<SCALARF:mode>_using_<FPCC:mode> swapped): To this. Add
CCFmode condition support.
(mov<mode>cc GPR): Expand R6 selects.
(mov<mode>cc FPR): Expand R6 selects.
(*tls_get_tp_<mode>_split): Do not .set push for >= mips32r2.
* config/mips/netbsd.h (TARGET_CPU_CPP_BUILTINS): Update similarly to
mips.h.
(ASM_SPEC): Add mips32r6, mips64r6.
* config/mips/t-isa3264 (MULTILIB_OPTIONS, MULTILIB_DIRNAMES): Update
for mips32r6/mips64r6.
* doc/invoke.texi: Document -mips32r6,-mips64r6.
* doc/md.texi: Update comment for ZD constraint.

libgcc/

* config.host: Support mipsisa32r6 and mipsisa64r6.
* config/mips/mips16.S: Do not build for R6.

gcc/testsuite/

* gcc.dg/torture/mips-hilo-2.c: Unconditionally pass for R6 onwards.
* gcc.dg/torture/pr19683-1.c: Likewise.
* gcc.target/mips/branch-cost-2.c: Require MOVN.
* gcc.target/mips/movcc-1.c: Likewise.
* gcc.target/mips/movcc-2.c: Likewise.
* gcc.target/mips/movcc-3.c: Likewise.
* gcc.target/mips/call-saved-4.c: Require LDC.
* gcc.target/mips/dmult-1.c: Require R5 or earlier.
* gcc.target/mips/fpcmp-1.c: Likewise.
* gcc.target/mips/fpcmp-2.c: Likewise.
* gcc.target/mips/neg-abs-2.c: Likewise.
* gcc.target/mips/timode-1.c: Likewise.
* gcc.target/mips/unaligned-1.c: Likewise.
* gcc.target/mips/madd-3.c: Require MADD.
* gcc.target/mips/madd-9.c: Likewise.
* gcc.target/mips/maddu-3.c: Likewise.
* gcc.target/mips/msub-3.c: Likewise.
* gcc.target/mips/msubu-3.c: Likewise.
* gcc.target/mips/mult-1.c: Require INS and not DMUL.
* gcc.target/mips/mips-ps-type-2.c: Require MADD.PS.
* gcc.target/mips/mips.exp (mips_option_groups): Add ins, dmul, ldc,
movn, madd, maddps.
(mips-dg-options): INS available from R2.  LDC available from MIPS II,
DMUL is present in octeon.  Describe all features removed from R6.

Co-Authored-By: Steve Ellcey <sellcey@imgtec.com>
From-SVN: r218973

Fix issue with string options and nested gcc_jit_contexts

gcc/jit/ChangeLog:
* jit-recording.c (gcc::jit::recording::context::context): When
copying string options from a parent context, take a copy of the
underlying buffers, rather than simply copying the pointer.

From-SVN: r218972

re PR target/64268 (bootstrap failure (ICE in wide_int_to_tree, at tree.c:1438) on powerpc in libgcc, stage2)

PR target/64268
* combine.c (try_combine): Immediately return if any of I0,I1,I2
are the same insn.

From-SVN: r218971

Handle NULL in gcc_jit_context_set_str_option

gcc/jit/ChangeLog:
* jit-recording.c (gcc::jit::recording::context::set_str_option):
Handle NULL.

From-SVN: r218969

re PR fortran/64209 ([OOP] runtime segfault with CLASS(*), INTENT(OUT) dummy argument)

2014-12-19  Janus Weil  <janus@gcc.gnu.org>

PR fortran/64209
* trans-expr.c (gfc_trans_class_array_init_assign): Check if _def_init
component is non-NULL.
(gfc_trans_class_init_assign): Ditto.

2014-12-19  Janus Weil  <janus@gcc.gnu.org>

PR fortran/64209
* gfortran.dg/unlimited_polymorphic_19.f90: New.

From-SVN: r218968

New std::string implementation.

* acinclude.m4 (GLIBCXX_ENABLE_LIBSTDCXX_CXX11_ABI): Remove.
(GLIBCXX_ENABLE_LIBSTDCXX_DUAL_ABI, GLIBCXX_DEFAULT_ABI): Add.
* configure.ac: Use new macros.
* configure: Regenerate.
* Makefile.in: Regenerate.
* doc/Makefile.in: Regenerate.
* libsupc++/Makefile.in: Regenerate.
* po/Makefile.in: Regenerate.
* src/Makefile.in: Regenerate.
* testsuite/Makefile.in: Regenerate.
* include/Makefile.am: Set _GLIBCXX_USE_DUAL_ABI.
* include/Makefile.in: Regenerate.
* config/abi/pre/gnu.ver: Export symbols related to new std::string.
Tighten old patterns to not match new symbols.
* config/locale/generic/monetary_members.cc: Guard some definitions
to not compile with new ABI.
* config/locale/gnu/monetary_members.cc: Likewise.
* config/locale/gnu/numeric_members.cc: Prevent double-free.
* config/os/gnu-linux/ldbl-extra.ver: Add new __gnu_cxx_ldbl128
exports. Tighten old patterns.
* doc/xml/manual/configure.xml: Document new configure options.
* doc/html/*: Regenerate.
* include/bits/basic_string.h (__cxx11::basic_string): Define new
non-reference-counted implementation in inline namespace __cxx11.
(stoi, stol, stoll, stof, stod, stold, to_string): Conditionally use
inline namespace.
(literals::string_literals::operator"): Conditionally use abi-tag.
* include/bits/basic_string.tcc (__cxx11::basic_string): Define.
* include/bits/c++config: Define _GLIBCXX_USE_DUAL_ABI and
LDBL_CXX11_ABI namespace macros.
* include/bits/locale_classes.h (locale::name()): Use abi_tag when
new ABI is in use.
(locale::_S_twinned_facets): New static member.
(locale::facet::__shim): Declare new type.
(locale::_facet::_M_sso_shim, locale::_facet::_M_cow_shim): New
functions for creating shims.
(locale::_Impl::_M_facet_unchecked): New member function for use
during construction.
(locale::_Impl::_M_init_extra): New member functions to create second
version of some facets.
(collate, collate_byname): Use abi_tag when new ABI is in use.
* include/bits/locale_facets.h: Add _GLIBCXX_NUM_CXX11_FACETS macro.
(numpunct, numpunct_byname): Use __cxx11 namespace.
(num_get::_M_extract_float, num_get::_M_extract_int): Use abi_tag
when new ABI is in use.
(num_get::__do_get, num_put::__do_put): Do not declare long double
compat functions for new ABI.
* include/bits/locale_facets.tcc (num_get, num_put): Use abi_tag on
definitions.
(numpunct, numpunct_byname): Qualify explicit instantiations.
* include/bits/locale_facets_nonio.h (time_get, time_get_byname,
moneypunct, moneypunct_byname, money_get, money_put, messages,
messages_byname): Use new inline namespace macros.
(money_get::__do_get, money_put::__do_put): Do not declare long
double compat functions for new ABI.
* include/bits/locale_facets_nonio.tcc (money_get, money_put): Use
new namespace macros.
(money_get::__do_get, money_put::__do_put): Do not define for new ABI.
* include/bits/localefwd.h (numpunct, numpunct_byname, collate,
collate_byname, time_get, time_get_byname, moneypunct,
moneypunct_byname, money_get, money_put, messages, messages_byname):
Use new namespace macros.
* include/bits/regex.h: Use inline namespace macros.
* include/bits/stl_list.h (_List_base, list): Use inline namespace
instead of abi-tag.
* include/bits/stringfwd.h (basic_string): Use namespace macros.
* include/std/iosfwd (basic_stringbuf, basic_istringstream,
basic_ostringstream, basic_stringstream): Likewise.
* include/std/sstream: Likewise.
(basic_stringbuf::__xfer_bufptrs): Update streambuf pointers on move.
* include/std/stdexcept (__cow_string, __sso_string): New types for
indirectly using std::string with either ABI.
(logic_error, runtime_error): Replace std::string member with
__cow_string when new ABI is in use. Declare non-inline copy
constructor and assignment operator. Declare const char* constructors.
(domain_error, invalid_argument, length_error, out_of_range,
range_error, overflow_error, underflow_error): Declare const char*
constructors.
* include/std/system_error (error_category): Replace with new
definition in inline namespace _V2.
(error_code::message, error_condition::message): Use abi_tag on
functions returning std::string.
* python/libstdcxx/v6/printers.py (StdStringPrinter): Handle new ABI.
* src/c++11/Makefile.am: Add new files.
* src/c++11/Makefile.in: Regenerate.
* src/c++11/compatibility-c++0x.cc: Compile with old std::string ABI.
Define old error_category symbols.
* src/c++11/cow-fstream-inst.cc: New. Instantiate fstream members
using old std::string ABI.
* src/c++11/cow-locale_init.cc (locale::_Impl::_M_init_extra): Define.
* src/c++11/cow-shim_facets.cc: Define shim facets using old ABI.
* src/c++11/cow-sstream-inst.cc: Instantiate stringstreams using old
std::string ABI.
* src/c++11/cow-stdexcept.cc: Define new constructors and assignment
operators.
(__cow_string, error_category::_M_message): Define.
* src/c++11/cow-string-inst.cc: Explicit instantiations using old
std::string. Include src/c++98/istream-string.cc.
* src/c++11/cow-wstring-inst.cc: Explicit instantiations using old
std::wstring.
* src/c++11/cxx11-hash_tr1.cc: Explicit instantiations using new
string.
* src/c++11/cxx11-ios_failure.cc: Add sanity check.
* src/c++11/cxx11-locale-inst.cc: Instantiate facets using new
std::string.
* src/c++11/cxx11-shim_facets.cc: Define shim facets using new ABI.
* src/c++11/cxx11-stdexcept.cc: Define constructors taking new
std::string.
* src/c++11/cxx11-wlocale-inst.cc: Instantiate facets using
new std::wstring.
* src/c++11/fstream-inst.cc: Compile with new ABI.
* src/c++11/functexcept.cc: Compile with old ABI.
* src/c++11/random.cc: Compile with new ABI.
* src/c++11/sstream-inst.cc: Compile with new ABI.
* src/c++11/string-inst.cc: Explicit instantiations for new string.
* src/c++11/system_error.cc (__sso_string, error_category::_M_message):
Define.
* src/c++11/wstring-inst.cc: Compile with new ABI.
* src/c++98/Makefile.am: Compile some host files twice for old and
new std::string. Add new files.
* src/c++98/Makefile.in: Regenerate.
* src/c++98/compatibility-ldbl.cc: Compile with old ABI.
* src/c++98/compatibility.cc: Likewise.
* src/c++98/concept-inst.cc: Likewise.
* src/c++98/hash_tr1.cc: Likewise.
* src/c++98/istream-string.cc: New file defining functions that
work with istream and std::string moved from ...
* src/c++98/istream.cc: ... here.
* src/c++98/cow-istream-string.cc: Recompile istream-string.cc with
old ABI.
* src/c++98/locale-inst.cc: Adjust facet instantiations to work for
either ABI.
* src/c++98/locale.cc (locale::_M_install_facet,
locale::_M_install_cache): Handle twinned facets.
* src/c++98/locale-facets.cc: Compile with old std::string ABI.
(__verify_grouping): Define new overload and old std::string version.
* src/c++98/locale_init.cc: Initialize twinned facets.
* src/c++98/localename.cc: Likewise.
* src/c++98/misc-inst.cc: Instantiate C++98-only std::string members.
(__verify_grouping): Define new std::string version.
* src/c++98/stdexcept.cc: Compile with old std::string ABI.
* src/c++98/wlocale-inst.cc: Likewise.
* testsuite/18_support/50594.cc: Adjust to work with SSO strings.
* testsuite/21_strings/basic_string/capacity/1.cc: Likewise.
* testsuite/21_strings/basic_string/capacity/char/1.cc: Likewise.
* testsuite/21_strings/basic_string/capacity/char/18654.cc: Likewise.
* testsuite/21_strings/basic_string/capacity/char/2.cc: Likewise.
* testsuite/21_strings/basic_string/capacity/wchar_t/1.cc: Likewise.
* testsuite/21_strings/basic_string/capacity/wchar_t/18654.cc:
Likewise.
* testsuite/21_strings/headers/string/synopsis.cc: Use inline
namespace macros.
* testsuite/23_containers/headers/list/synopsis.cc: Likewise.
* testsuite/27_io/basic_ios/copyfmt/char/1.cc: Set dg-options so
correct exception type can be caught.
* testsuite/27_io/basic_ios/exceptions/char/1.cc: Likewise.
* testsuite/27_io/basic_istream/extractors_arithmetic/char/
exceptions_failbit.cc: Likewise.
* testsuite/27_io/basic_istream/extractors_arithmetic/wchar_t/
exceptions_failbit.cc: Likewise.
* testsuite/27_io/basic_istream/extractors_other/char/
exceptions_null.cc: Likewise.
* testsuite/27_io/basic_istream/extractors_other/wchar_t/
exceptions_null.cc: Likewise.
* testsuite/27_io/basic_istream/sentry/char/12297.cc: Likewise.
* testsuite/27_io/basic_istream/sentry/wchar_t/12297.cc: Likewise.
* testsuite/27_io/basic_ostream/inserters_other/char/
exceptions_null.cc: Likewise.
* testsuite/27_io/basic_ostream/inserters_other/wchar_t/
exceptions_null.cc: Likewise.
* testsuite/27_io/ios_base/storage/2.cc: Likewise.
* testsuite/27_io/ios_base/failure/cxx11.cc: Disable for old ABI.
* testsuite/ext/profile/mutex_extensions_neg.cc: Adjust dg-error.
* testsuite/libstdc++-prettyprinters/libfundts.cc: Use old ABI.
* testsuite/libstdc++-prettyprinters/simple.cc: Likewise.
* testsuite/libstdc++-prettyprinters/simple11.cc: Likewise.
* testsuite/libstdc++-prettyprinters/whatis.cc: Likewise.
* testsuite/util/exception/safety.h: Adjust member function types
for new std::string.
* testsuite/util/testsuite_abi.cc: Add new version and ignore
__float128 symbols in __cxx11 namespace.

From-SVN: r218964

2014-12-19  Paolo Carlini  <paolo.carlini@oracle.com>

* Tweak ChangeLog.

From-SVN: r218963

using30.C: Move ...

        * g++.dg/template/using30.C: Move ...
        * g++.dg/cpp0x/alias-decl-45.C: ... here.

From-SVN: r218962

[AArch64 3/3] Fix XOR_one_cmpl pattern; add SIMD-reg variants for BIC,ORN,EON

gcc/:

* config/aarch64/aarch64.c (<LOGICAL:optab>_one_cmpl<mode>3):
Reparameterize to...
(<NLOGICAL:optab>_one_cmpl<mode>3): with extra SIMD-register variant.
(xor_one_cmpl<mode>3): New define_insn_and_split.

* config/aarch64/iterators.md (NLOGICAL): New define_code_iterator.

gcc/testsuite/:

* gcc.target/aarch64/eon_1.c: New test.

From-SVN: r218961

[AArch64 2/3] Add SIMD-reg variants of logical operators and/ior/xor/not

* config/aarch64/aarch64.md (<optab><mode>3, one_cmpl<mode>2):
Add SIMD-register variant.
* config/aarch64/iterators.md (Vbtype): Add value for SI.

From-SVN: r218960

using30.C: Move...

2014-12-19  Paolo Carlini  <paolo.carlini@oracle.com>

* g++.dg/template/using30.C: Move...
* g++.dg/cpp0x/alias-decl-45.C: ... here.

From-SVN: r218959

[AArch64 1/3] Don't disparage add/sub in SIMD registers

        * config/aarch64/aarch64.md (subdi3, adddi3_aarch64): Don't penalize
        SIMD reg variant.

From-SVN: r218958

re PR c++/60493 (g++ throws segmentation fault on simple code)

2014-12-19  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/60493
* g++.dg/template/crash120.C: New.

From-SVN: r218957

re PR c++/61198 (Crash when selecting specializations through aliases.)

PR c++/61198
* g++.dg/template/using30.C: New file.

From-SVN: r218956

re PR c++/61198 (Crash when selecting specializations through aliases.)

PR c++/61198
* pt.c (most_general_template): Don't break for template-alias.

From-SVN: r218955

compiler: More cases that need a temporary for interface conversion.

From-SVN: r218952

re PR ipa/63569 (Wrong code with volatile and ICF)

Fix for PR ipa/63569.

PR ipa/63569
* gcc.dg/ipa/pr63569.c: New test.
PR ipa/63569
* ipa-icf-gimple.c (func_checker::compare_operand): Add missing
comparison for volatile flag.

From-SVN: r218949

re PR preprocessor/63831 (r217292 causes segfaults with -MM)

PR preprocessor/63831
* c-cppbuiltin.c (c_cpp_builtins): Don't define __has_attribute
and __has_cpp_attribute here.
* c-ppoutput.c (init_pp_output): Set cb->has_attribute to
c_common_has_attribute.
* c-common.h (c_common_has_attribute): New prototype.
* c-lex.c (init_c_lex): Set cb->has_attribute to
c_common_has_attribute instead of cb_has_attribute.
(get_token_no_padding): New function.
(cb_has_attribute): Renamed to ...
(c_common_has_attribute): ... this.  No longer static.  Use
get_token_no_padding, require ()s, don't build TREE_LIST
unnecessarily, fix up formatting, adjust diagnostics, call
init_attributes.

* directives.c (lex_macro_node): Remove __has_attribute__ handling.
* internal.h (struct spec_node): Remove n__has_attribute__ field.
(struct lexer_state): Remove in__has_attribute__ field.
* macro.c (_cpp_builtin_macro_text): Handle BT_HAS_ATTRIBUTE.
* identifiers.c (_cpp_init_hashtable): Remove __has_attribute__
handling.
* init.c (builtin_array): Add __has_attribute and __has_cpp_attribute.
(cpp_init_special_builtins): Don't initialize __has_attribute
or __has_cpp_attribute if CLK_ASM or pfile->cb.has_attribute is NULL.
* traditional.c (enum ls): Remove ls_has_attribute,
ls_has_attribute_close.
(_cpp_scan_out_logical_line): Remove __has_attribute__ handling.
* include/cpplib.h (enum cpp_builtin_type): Add BT_HAS_ATTRIBUTE.
* pch.c (cpp_read_state): Remove __has_attribute__ handling.
* expr.c (eval_token): Likewise.
(parse_has_attribute): Removed.

* c-c++-common/cpp/pr63831-1.c: New test.
* c-c++-common/cpp/pr63831-2.c: New test.

From-SVN: r218948

[ARM][doc] Remove mention of Advanced RISC Machines

* doc/invoke.texi (ARM options): Remove mention of Advanced RISC
Machines.

From-SVN: r218897

Fix ChangeLog typo in 218895

From-SVN: r218896

[PATCH][ARM] Fix reservation pattern in cortex-a9-neon.md

gcc/

* config/arm/cortex-a9-neon.md (cortex_a9_neon_vmov): Change
reservation to cortex_a9_neon_dp.

From-SVN: r218895

* [SH] Split QI/HImode load/store via r0 when LRA is enabled.

From-SVN: r218892

* [SH] Add splitter to addsi3_compact.

From-SVN: r218891

* [SH] Modify movsi_ie and movsf_ie patterns for LRA.

From-SVN: r218890

* [SH] Miscellaneous changes for LRA.

From-SVN: r218889

* [SH] Add -mlra option.

From-SVN: r218888

* Add TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT target macro.

From-SVN: r218887

* Add TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV target macro.

From-SVN: r218886

lra-constraints.c (process_address_1): Swap base_term and index_term if INDEX_REG_CLASS is assigned to base_term...

* lra-constraints.c (process_address_1): Swap base_term and index_term
  if INDEX_REG_CLASS is assigned to base_term already when INDEX_REG_CLASS
  is a single register class.

From-SVN: r218885

compiler: Avoid multiple evaluations in interface conversions.

Added assertions for cases that might lead to multiple
evaluations, and fixed all the problems I saw.

Test case already in master Go testsuite
(https://go-review.googlesource.com/#/c/1710/).

From-SVN: r218884

Daily bump.

From-SVN: r218883

re PR c++/63723 (Narrowing conversion allowed in braced init list in SFINAE context)

2014-12-18  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/63723
* g++.dg/cpp0x/sfinae54.C: New.

From-SVN: r218880

re PR c++/64105 (ICE: in strip_typedefs, at cp/tree.c:1326)

PR c++/64105
* parser.c (cp_parser_simple_type_specifier): Make auto parameter
before -std=c++14 an error.

From-SVN: r218879

re PR c++/59204 (Incorrect metaprogram evaluation in SFINAE context)

2014-12-18  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/59204
* g++.dg/cpp0x/sfinae53.C: New.

From-SVN: r218878

re PR c++/64352 (No SFINAE with deleted function)

PR c++/64352
* pt.c (tsubst_copy_and_build): Pass complain to mark_used.

From-SVN: r218877

re PR c++/64251 (Chromium build error only with --enable-checking=yes)

PR c++/64251
* decl2.c (mark_used): Don't mark if in_template_function.

From-SVN: r218876

lra-constraints.c (lra-constraints.c): Exchange places of sclass and dclass.

2014-12-18  Vladimir Makarov  <vmakarov@redhat.com>

* lra-constraints.c (lra-constraints.c): Exchange places of sclass
and dclass.

From-SVN: r218875

re PR rtl-optimization/64291 (Miscompile t-div in GMP's testsuite)

2014-12-18  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/64291
* lra-remat.c (bad_for_rematerialization_p): Add UNPSEC_VLOATILE.
(create_cands): Process only output reload insn with potential
cands.

2014-12-18  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/64291
* testsuite/gcc.target/i386/pr64291-[12].c: New tests.

From-SVN: r218874

re PR c++/60955 (Erroneous warning about taking address of register with std=c++1y)

/cp
2014-12-18  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/60955
* pt.c (struct warning_sentinel): Move it...
* cp-tree.h: ... here.
* semantics.c (force_paren_expr): Use it.

/testsuite
2014-12-18  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/60955
* g++.dg/warn/register-parm-1.C: New.

From-SVN: r218871

X86-64: Add -mskip-rax-setup

The Linux kernel never passes floating point arguments around, vararg
functions or not. Hence no vector registers are ever used when calling a
vararg function.  But gcc still dutifully emits an "xor %eax,%eax" before
each and every call of a vararg function.  Since no callee use that for
anything, these instructions are redundant.

This patch adds the -mskip-rax-setup option to skip setting up RAX
register when SSE is disabled and there are no variable arguments passed
in vector registers.  Since RAX register is used to avoid unnecessarily
saving vector registers on stack when passing variable arguments, the
impacts of this option are callees may waste some stack space, misbehave
or jump to a random location.  GCC 4.4 or newer don't those issues,
regardless the RAX register value since they don't check the RAX register
value when SSE is disabled.

gcc/

* config/i386/i386.c (ix86_expand_call): Skip setting up RAX
register for -mskip-rax-setup when there are no parameters
passed in vector registers.
* config/i386/i386.opt (mskip-rax-setup): New option.
* doc/invoke.texi: Document -mskip-rax-setup.

gcc/testsuite/

* gcc.target/i386/amd64-abi-7.c: New tests.
* gcc.target/i386/amd64-abi-8.c: Likwise.
* gcc.target/i386/amd64-abi-9.c: Likwise.

From-SVN: r218870

[AArch64] Simplify+improve patterns for ushr(d?)_n_u64 intrinsic

gcc/:
* config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): Handle shift
by 64 by moving const0_rtx.
(aarch64_ushr_simddi): Delete.

* config/aarch64/aarch64.md (enum unspec): Delete UNSPEC_USHR64.

gcc/testsuite/:

        * gcc.target/aarch64/ushr64_1.c: Remove scan-assembler "ushr...64".

From-SVN: r218868

[AArch64] Simplify patterns for sshr_n_[us]64 intrinsic

        * config/aarch64/aarch64.md (enum "unspec"): Remove UNSPEC_SSHR64.

        * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): Change shift
        amount to 63 if was 64.
        (aarch64_sshr_simddi): Remove.

From-SVN: r218867

[AArch64] Add TARGET_MIN_DIVISIONS_FOR_RECIP_MUL

2014-12-18  Wilco Dijkstra  <wilco.dijkstra@arm.com>

* gcc/config/aarch64/aarch64.c (TARGET_MIN_DIVISIONS_FOR_RECIP_MUL):
Define.
(aarch64_min_divisions_for_recip_mul): New function.

From-SVN: r218866

[AArch64] Generalize code alignment

2014-12-18  Wilco Dijkstra  <wilco.dijkstra@arm.com>

* config/aarch64/aarch64-protos.h (tune-params): Add code alignment
tuning parameters.
* gcc/config/aarch64/aarch64.c (generic_tunings): Add code alignment
tuning parameters.
(cortexa53_tunings): Likewise.
(cortexa57_tunings): Likewise.
(thunderx_tunings): Likewise.
(aarch64_override_options): Use new alignment tunings.

From-SVN: r218865

re PR tree-optimization/64330 (IPA-ICF merges const exported vars that could be addressable in other TUs)

Fix for PR64330.

PR tree-optimization/64330
* ipa-icf.c (sem_variable::parse): Add checking
for externally visible symbols and do not introduce
an alias for an external declaration.

From-SVN: r218864

re PR bootstrap/63573 (libgo: ICE building libgo on powerpc-linux-gnu)

Fix for PR bootstrap/63573.

PR bootstrap/63573
* tree-inline.c (remap_gimple_stmt): Handle gimple_call_from_thunk_p
predicate.

From-SVN: r218861

re PR ipa/64146 (ipa-icf breaks gold dynamic_test_2 test)

Fix for PR ipa/64146

PR ipa/64146
* ipa-icf.c (sem_function::merge): Check for
decl_binds_to_current_def_p is newly added to merge operation.
* g++.dg/ipa/pr64146.C: New test.

From-SVN: r218860

Change mpx effective-target test into link test

* lib/mpx-dg.exp (check_effective_target_mpx): Change into link test.
Add main.

From-SVN: r218859

Include target-utils.exp in boehm-gc testing

* testsuite/lib/boehm-gc.exp: Load target-utils.exp.

From-SVN: r218857

Add myself to MAINTAINERS under Writer After Approval

From-SVN: r218856

re PR middle-end/62178 ([AArch64] Performance regression on matrix matrix multiply due to r211211)

PR tree-optimization/62178
* tree-ssa-loop-ivopts.c (cheaper_cost_with_cand): New function.
(iv_ca_replace): New function.
(try_improve_iv_set): New parameter try_replace_p.
Break local optimal fixed-point by calling iv_ca_replace.
(find_optimal_iv_set_1): Pass new argument to try_improve_iv_set.

gcc/testsuite/ChangeLog
PR tree-optimization/62178
* gcc.target/aarch64/pr62178.c: New test.

From-SVN: r218855

auto-profile.c (afdo_annotate_cfg): Invoke update_ssa in the right place.

gcc/ChangeLog:
2014-12-17  Dehao Chen  <dehao@google.com>

        * auto-profile.c (afdo_annotate_cfg): Invoke update_ssa in the right
        place.
        (auto_profile): Recompute inline summary after processing cgraph node.

From-SVN: r218854

Daily bump.

From-SVN: r218853

re PR target/51244 ([SH] Inefficient conditional branch and code around T bit)

gcc/
PR target/51244
* config/sh/sh_treg_combine.cc (is_conditional_insn): New function.
(cbranch_trace): Add member rtx* condition_rtx_in_insn, initialize it
accordingly in constructor.
(cbranch_trace::branch_condition_rtx_ref): New function.
(cbranch_trace::branch_condition_rtx): Use branch_condition_rtx_ref.
(sh_treg_combine::try_invert_branch_condition): Invert condition rtx
in insn using reversed_comparison_code and validate_change instead of
invert_jump_1.
(sh_treg_combine::execute): Look for conditional insns in basic blocks
in addition to conditional branches.
* config/sh/sh.md (*movsicc_div0s): Remove combine patterns.

From-SVN: r218850

re PR target/51244 ([SH] Inefficient conditional branch and code around T bit)

gcc/
PR target/51244
* config/sh/sh_treg_combine.cc (sh_treg_combine::try_optimize_cbranch):
Combine ccreg inversion and cbranch into inverted cbranch.

From-SVN: r218847

dg-extract-results.sh: Use --text with grep to avoid issues with binary files.

2014-12-17  Sergio Durigan Junior  <sergiodj@redhat.com>

* dg-extract-results.sh: Use --text with grep to avoid issues with
  binary files.  Fall back to cat -v, if that doesn't work.

From-SVN: r218843