Daily bump.

From-SVN: r222438

mips.md: (*madd4<mode>) Remove accum_in attribute.

2015-04-24  Steve Ellcey  <sellcey@imgtec.com>

* config/mips/mips.md: (*madd4<mode>) Remove accum_in attribute.
(*madd3<mode>): Ditto.
(*msub4<mode>): Ditto.
(*msub3<mode>): Ditto.
(*nmadd4<mode>): Ditto.
(*nmadd3<mode>): Ditto.
(*nmadd4<mode>_fastmath): Ditto.
(*nmadd3<mode>_fastmath): Ditto.
(*nmsub4<mode>): Ditto.
(*nmsub3<mode>): Ditto.
(*nmsub4<mode>_fastmath): Ditto.
(*nmsub3<mode>_fastmath): Ditto.

From-SVN: r222432

re PR go/65616 (libgobegin.a not compiled with -fPIC)

PR go/65616
libgo: Compile go-main, in libgobegin, with -fPIC.

From-SVN: r222431

re PR c++/50800 (Internal compiler error in finish_member_declarations, possibly related to may_alias attribute)

PR c++/50800
* g++.dg/ext/alias-canon2.C: Remove clashing fn definitions.

From-SVN: r222426

re PR c++/50800 (Internal compiler error in finish_member_declarations, possibly related to may_alias attribute)

PR c++/50800
gcc/
* tree.c (build_reference_type_for_mode): Don't pass can_alias_all
down when building TYPE_CANONICAL.
(build_pointer_type_for_mode): Likewise.
gcc/cp/
* typeck.c (structural_comptypes): Don't check TYPE_REF_CAN_ALIAS_ALL.

From-SVN: r222419

genrecog.c (validate_pattern): Check matching constraint refers to a lower numbered operand.

        * genrecog.c (validate_pattern): Check matching constraint refers
        to a lower numbered operand.

From-SVN: r222418

re PR target/65849 (Add missing tuning knobs to #pragma target/attribute target support)

[gcc]
2015-04-24  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/65849
* config/rs6000/rs6000.opt (-mvsx-align-128): Make options that
save to independent variables use the Save attribute.  This will
allow these options to be modified with the #pragma/attribute
target support.
(-mallow-movmisalign): Likewise.
(-mallow-df-permute): Likewise.
(-msched-groups): Likewise.
(-malways-hint): Likewise.
(-malign-branch-targets): Likewise.
(-mvectorize-builtins): Likewise.
(-msave-toc-indirect): Likewise.

* config/rs6000/rs6000.c (rs6000_opt_masks): Add more options that
can be set via the #pragma/attribute target support.
(rs6000_opt_vars): Likewise.
(rs6000_inner_target_options): If VSX was set, also set
-mno-avoid-indexed-addresses.

[gcc/testsuite]
2015-04-24  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/65849
* gcc.target/powerpc/pr65849-1.c: New test to verify being able to
set new options.
* gcc.target/powerpc/pr65849-2.c: Likewise.

From-SVN: r222417

[ARM][trivial] Use uppercase for code iterator names

* config/arm/iterators.md (shiftable_ops): Rename to...
(SHIFTABLE_OPS): ... This.  Update use in comments.
(ior_xor): Rename to...
(IOR_XOR): ... This.
(vqh_ops): Rename to...
(VQH_OPS): ... This.
(vqhs_ops): Rename to...
(VQHS_OPS): ... This.
(rshifts): Rename to...
(RSHIFTS): ... This.
(returns): Rename to...
(RETURNS): ... This.
* config/arm/arm.md: Update uses of the above.
* config/arm/neon.md: Likewise.

From-SVN: r222416

[AArch64] Implement -m{cpu,tune,arch}=native using only /proc/cpuinfo

* config.host (case ${host}): Add aarch64*-*-linux case.
* config/aarch64/aarch64-cores.def: Add IMPLEMENTER_ID and PART_NUMBER
fields to all the cores.
* config/aarch64/aarch64-elf.h (DRIVER_SELF_SPECS):
Add MCPU_MTUNE_NATIVE_SPECS.
* config/aarch64/aarch64-option-extensions.def: Add FEATURE_STRING
field to all extensions.
* config/aarch64/aarch64-opts.h: Adjust definition of AARCH64_CORE.
* config/aarch64/aarch64.c: Adjust definition of AARCH64_CORE.
Adjust definition of AARCH64_OPT_EXTENSION.
* config/aarch64/aarch64.h: Adjust definition of AARCH64_CORE.
(MCPU_MTUNE_NATIVE_SPECS): Define.
* config/aarch64/driver-aarch64.c: New file.
* config/aarch64/x-arch64: New file.
* doc/invoke.texi (AArch64 Options): Document native value for -mcpu,
-mtune and -march.

From-SVN: r222415

constexpr.c (potential_constant_expression_1): Remove obsolete code.

* constexpr.c (potential_constant_expression_1) [MINUS_EXPR]:
Remove obsolete code.
[NE_EXPR]: Likewise.

From-SVN: r222414

Replace g++.dg/pr65802.C with gcc.dg/pr65802.c

2015-04-24  Tom de Vries  <tom@codesourcery.com>

PR tree-optimization/65802
* g++.dg/pr65802.C: Move to ...
* gcc.dg/pr65802.c: ... here.  Add -fexceptions to dg-options. Include
stdarg.h.  Rewrite for C.
(fn1): Use va_list and va_arg.  Make variable args function.  Add use of
va_start and va_end.  Remove unnecessary inline asm.

From-SVN: r222413

i386-protos.h (ix86_operands_ok_for_move_multiple): New.

* config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple): New.
* config/i386/i386.c (extract_base_offset_in_addr): New function.
(ix86_operands_ok_for_move_multiple): Ditto.
* config/i386/sse.md (movsd/movhpd to movupd peephole2): New pattern.
(movlpd/movhpd to movupd peephole2): Ditto.

testsuite/ChangeLog:

* gcc.target/i386/sse2-load-multi.c: New test.
* gcc.target/i386/sse2-store-multi.c: Ditto.

Co-Authored-By: Wei Mi <wmi@google.com>
From-SVN: r222410

re PR c/63357 (Warn for P && P and P || P  (same expression used multiple times in a condition))

PR c/63357
* c-common.c (warn_logical_operator): Warn if the operands have the
same expressions.

* doc/invoke.texi: Update description of -Wlogical-op.

* c-c++-common/Wlogical-op-1.c: New test.

From-SVN: r222408

re PR c/65830 (-Wno-shift-count-negative -Wno-shift-count-overflow don't work with const ints)

PR c/65830
* c-common.c (c_fully_fold_internal): Use OPT_Wshift_count_negative
and OPT_Wshift_count_overflow.

* c-c++-common/pr65830.c: New test.

From-SVN: r222407

re PR c/61534 (Wlogical-op should not warn when either operand comes from macro expansion)

PR c/61534
* input.h (from_macro_expansion_at): Define.

* c-common.c (warn_logical_operator): Bail if either operand comes
from a macro expansion.

* c-c++-common/pr61534-1.c: New test.

From-SVN: r222406

sse.md (*vec_widen_smult_even_v8si<mask_name>): Mark operand1 commutative.

* config/i386/sse.md (*vec_widen_smult_even_v8si<mask_name>):
Mark operand1 commutative.

From-SVN: r222403

unknown-elf.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): fix ternary operator in fprintf and harmonize spacing.

2015-04-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>

    * config/arm/unknown-elf.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): fix
    ternary operator in fprintf and harmonize spacing.

From-SVN: r222401

sse.md (*vec_concatv2sf_sse4_1): Do not allow both input operands in memory.

* config/i386/sse.md (*vec_concatv2sf_sse4_1): Do not allow both
input operands in memory.
(*vec_concatv2si_sse4_1): Ditto.
(*vec_concatv2df): Ditto, except for SSE3 and equal input operands.
(vec_extract_lo_<mode><mask_name>): Change operand 1 predicate to
register_operand.
(vec_extract_hi_v32hi): Ditto.
(vec_extract_hi_v64hi): Ditto.
(<mask_codefor>avx512f_unpckhpd512<mask_name>): Ditto.

From-SVN: r222400

re PR rtl-optimization/34503 (Issues with constant/copy propagation implementation in gcse.c)

2015-04-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
            Steven Bosscher <steven@gcc.gnu.org>

    gcc/
    PR rtl-optimization/34503
    * cprop.c (cprop_reg_p): New.
    (hash_scan_set): Use above function to check if register can be
    propagated.
    (find_avail_set): Return up to two sets, one whose source is a
    register and one whose source is a constant.  Sets are returned in an
    array passed as parameter rather than as a return value.
    (cprop_insn): Use a do while loop rather than a goto.  Try each of the
    sets returned by find_avail_set, starting with the one whose source is
    a constant. Use cprop_reg_p to check if register can be propagated.
    (do_local_cprop): Use cprop_reg_p to check if register can be
    propagated.
    (implicit_set_cond_p): Likewise.

    gcc/testsuite/
    PR rtl-optimization/34503
    * gcc.target/arm/pr64616.c: New file.

Co-Authored-By: Steven Bosscher <steven@gcc.gnu.org>
From-SVN: r222398

pr42172-1.c: Check str instead of ldr.

gcc/testsuite/ChangeLog
* gcc.target/arm/pr42172-1.c: Check str instead of ldr.

From-SVN: r222396

pr65710.c: Update the options.

2015-04-24  Terry Guo  <terry.guo@arm.com>

* gcc.target/arm/pr65710.c: Update the options.

From-SVN: r222393

ipa-icf.c (sem_function::equals_wpa): Compare thunk info.

* ipa-icf.c (sem_function::equals_wpa): Compare thunk info.
(sem_function::equals): IGNORED_NODES parameter is now unused;
update call of equals_private.
(sem_function::equals_private): Do not call equals_wpa; skip
gimple body matching if there is no body.
(sem_function::init): Add logic to hash tthunk info.
(sem_function::parse): Also parse thunks.
* ipa-icf.h (equals_private): Update declaration.

From-SVN: r222392

Daily bump.

From-SVN: r222391

altivec.md (*altivec_lvx_<mode>_internal): Remove asterisk from name so this can be generated directly.

2015-04-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/altivec.md (*altivec_lvx_<mode>_internal): Remove
asterisk from name so this can be generated directly.
(*altivec_stvx_<mode>_internal): Likewise.
* config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Add assert
that this is never called during or after reload/lra.
(rs6000_frame_related): Remove split_reg
argument and logic that references it.
(emit_frame_save): Remove last parameter from call to
rs6000_frame_related.
(rs6000_emit_prologue): Remove last parameter from eight calls to
rs6000_frame_related.  Force generation of stvx instruction for
Altivec register saves.  Remove split_reg handling, which is no
longer needed.
(rs6000_emit_epilogue):  Force generation of lvx instruction for
Altivec register restores.

From-SVN: r222385

Revert:PR c++/50800

Revert:PR c++/50800
       * tree.c (strip_typedefs): Add remove_attributes parm.
       (strip_typedefs_expr): Likewise.
       (apply_identity_attributes): New subroutine of strip_typedefs.
       * pt.c (canonicalize_type_argument): Let strip_typedefs handle attrs.
       (convert_nontype_argument, unify): Likewise.
       * cp-tree.h: Adjust.

From-SVN: r222384

Skip gcc.target/i386/pr65523.c for x32

* gcc.target/i386/pr65523.c: Skip x32.

From-SVN: r222382

rs6000.opt (mcrypto): Change option description to match category changes in ISA 2.07B.

2015-04-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/rs6000.opt (mcrypto): Change option description to
match category changes in ISA 2.07B.

From-SVN: r222381

tree.h (attribute_value_equal): Declare.

* tree.h (attribute_value_equal): Declare.
* tree.c (attribute_value_equal): Export.

From-SVN: r222380

[ARM] Rewrite vc<cond> NEON patterns to use RTL operations rather than UNSPECs

* config/arm/iterators.md (GTGE, GTUGEU, COMPARISONS): New code
iterators.
(cmp_op, cmp_type): New code attributes.
(NEON_VCMP, NEON_VACMP): New int iterators.
(cmp_op_unsp): New int attribute.
* config/arm/neon.md (neon_vc<cmp_op><mode>): New define_expand.
(neon_vceq<mode>): Delete.
(neon_vc<cmp_op><mode>_insn): New pattern.
(neon_vc<cmp_op_unsp><mode>_insn_unspec): Likewise.
(neon_vcgeu<mode>): Delete.
(neon_vcle<mode>): Likewise.
(neon_vclt<mode>: Likewise.
(neon_vcage<mode>): Likewise.
(neon_vcagt<mode>): Likewise.
(neon_vca<cmp_op><mode>): New define_expand.
(neon_vca<cmp_op><mode>_insn): New pattern.
(neon_vca<cmp_op_unsp><mode>_insn_unspec): Likewise.

* gcc.target/arm/neon/pr51534.c: Update vcg* scan-assembly patterns
to look for vcl* where appropriate.

From-SVN: r222379

Fix testuite for updated ICF dumps.

From-SVN: r222378

re PR c++/50800 (Internal compiler error in finish_member_declarations, possibly related to may_alias attribute)

PR c++/50800
* tree.c (strip_typedefs): Add remove_attributes parm.
(strip_typedefs_expr): Likewise.
(apply_identity_attributes): New subroutine of strip_typedefs.
* pt.c (canonicalize_type_argument): Let strip_typedefs handle attrs.
(convert_nontype_argument, unify): Likewise.
* cp-tree.h: Adjust.

From-SVN: r222377

re PR c++/65646 (ICE in invalid syntax)

PR c++/65646
* pt.c (check_explicit_specialization): Don't
SET_DECL_TEMPLATE_SPECIALIZATION for a variable with no template
headers.
* decl.c (grokvardecl): Revert earlier fix.

From-SVN: r222376

Correct name and email address in the Changelog commit.

From-SVN: r222375

ipa-icf.c (sem_item::compare_attributes): New function.

* ipa-icf.c (sem_item::compare_attributes): New function.
(sem_item::compare_referenced_symbol_properties): Compare variable
attributes.
(sem_item::hash_referenced_symbol_properties): Record DECL_ALIGN.
(sem_function::param_used_p): New function.
(sem_function::equals_wpa): Fix attribute comparsion; match
parameter type codes; do not compare paremter flags when
they are not used; compare edge flags; compare indirect calls.
(sem_item::update_hash_by_addr_refs): Hash reference type.
(sem_function::equals_private): Do not match DECL_ATTRIBUTES.
(sem_variable::equals_wpa): Do not match DECL_ALIGN; match
reference use type.
(sem_item_optimizer::update_hash_by_addr_refs): Use param_used_p.
* ipa-icf.h (compare_attributes, param_used_p): Declare.

From-SVN: r222374

costmodel-bb-slp-9a.c: Replace vect_no_align with vect_no_align && { ! vect_hw_misalign }.

2015-04-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gcc.dg/vect/costmodel/ppc/costmodel-bb-slp-9a.c: Replace
vect_no_align with vect_no_align && { ! vect_hw_misalign }.

From-SVN: r222373

ipa-icf.c (symbol_compare_collection::symbol_compare_collection): cleanup.

* ipa-icf.c (symbol_compare_collection::symbol_compare_collection):
cleanup.
(sem_function::get_hash): Do not hash DECL_DISREGARD_INLINE_LIMITS,
DECL_DECLARED_INLINE_P and DECL_IS_OPERATOR_NEW.
(sem_item::compare_referenced_symbol_properties): New.
(sem_item::hash_referenced_symbol_properties): New.
(sem_item::compare_cgraph_references): Rename to ...
(sem_item::compare_symbol_references): ... this one; use
compare_referenced_symbol_properties.
(sem_function::equals_wpa): Do not compare
DECL_DISREGARD_INLINE_LIMITS, DECL_DECLARED_INLINE_P,
DECL_IS_OPERATOR_NEW; compare pointer sizes.
(sem_item::update_hash_by_addr_refs): Call
hash_referenced_symbol_properties.
(sem_item::update_hash_by_local_refs): Cleanup.
(sem_function::merge): Do not mix up symbol properties.
(sem_variable::equals_wpa): Use compare_symbol_references.
* ipa-icf.h (sem_item::compare_referenced_symbol_properties): New.
(sem_item::hash_referenced_symbol_properties): New.
(sem_item::compare_symbol_references): New.
(sem_item::compare_cgraph_references): Remove.

From-SVN: r222372

re PR target/26702 (.size is not emitted for BSS variables)

Fix PR target/26702

For Kwok Cheung Yeung.

From-SVN: r222371

re PR c/65345 (ICE with _Generic selection on _Atomic int)

PR c/65345
* c-decl.c (set_labels_context_r): New function.
(store_parm_decls): Call it via walk_tree_without_duplicates.
* c-typeck.c (convert_lvalue_to_rvalue): Use create_tmp_var_raw
instead of create_tmp_var.  Build TARGET_EXPR instead of
COMPOUND_EXPR.
(build_atomic_assign): Use create_tmp_var_raw instead of
create_tmp_var.  Build TARGET_EXPRs instead of MODIFY_EXPR.

* gcc.dg/pr65345-1.c: New test.
* gcc.dg/pr65345-2.c: New test.

From-SVN: r222370

rl78.c (rl78_preferred_reload_class): Add ATTRIBUTE_UNUSED to x parameter.

* config/rl78/rl78.c (rl78_preferred_reload_class): Add
ATTRIBUTE_UNUSED to x parameter.
* config/rl78/rl78-opts.h (enum rl78_mul_types): Remove unused MUL_RL78.

From-SVN: r222369

crypto.md (crypto_vpmsum<CR_char>): Change TARGET_CRYPTO to TARGET_P8_VECTOR>

[gcc]

2015-04-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/crypto.md (crypto_vpmsum<CR_char>): Change
TARGET_CRYPTO to TARGET_P8_VECTOR>
(crypto_vpermxor_<mode>): Likewise.
* config/rs6000/rs6000-builtin.def (BU_CRYPTO_2A): New #define.
(BU_CRYPTO_3A): Likewise.
(BU_CRYPTO_OVERLOAD_2A): Rename from BU_CRYPTO_OVERLOAD_2.
(BU_CRYPTO_OVERLOAD_3A): New #define.
(VPMSUMB): Change from BU_CRYPTO_2 to BU_CRYPTO_2A.
(VPMSUMH): Likewise.
(VPMSUMW): Likewise.
(VPMSUMD): Likewise.
(VPERMXOR_V2DI): Change from BU_CRYPTO_3 to BU_CRYPTO_3A.
(VPERMXOR_V4SI): Likewise.
(VPERMXOR_V8HI): Likewise.
(VPERMXOR_V16QI): Likewise.
(VPMSUM): Change from BU_CRYPTO_OVERLOAD_2 to
BU_CRYPTO_OVERLOAD_2A.
(VPERMXOR): Change from BU_CRYPTO_OVERLOAD3 to
BU_CRYPTO_OVERLOAD_3A.
* config/rs6000/rs6000.opt (mcrypto): Change description of
option.

[gcc/testsuite]

2015-04-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gcc.target/powerpc/crypto-builtin-2.c: New.

From-SVN: r222362

PF fortran/60322

gcc/testsuite/ChangeLog:

2015-04-23  Andre Vehreschild  <vehre@gmx.de>

PF fortran/60322
* gfortran.dg/class_allocate_19.f03: New test.
* gfortran.dg/class_array_20.f03: New test.
* gfortran.dg/class_array_21.f03: New test.
* gfortran.dg/finalize_10.f90: Corrected scan-trees.
* gfortran.dg/finalize_15.f90: Fixing comparision to model
initialization correctly.
* gfortran.dg/finalize_29.f08: New test.

gcc/fortran/ChangeLog:

2015-04-23  Andre Vehreschild  <vehre@gmx.de>

PR fortran/60322
* expr.c (gfc_lval_expr_from_sym): Code to select the regular
or class array added.
* gfortran.h: Add IS_CLASS_ARRAY macro.
* trans-array.c (gfc_add_loop_ss_code): Treat class objects
to be referenced always.
(build_class_array_ref): Adapt retrieval of array descriptor.
(build_array_ref): Likewise.
(gfc_conv_array_ref): Hand the vptr or the descriptor to
build_array_ref depending whether the sym is class or not.
(gfc_trans_array_cobounds):  Select correct gfc_array_spec for
regular and class arrays.
(gfc_trans_array_bounds): Likewise.
(gfc_trans_dummy_array_bias): Likewise.
(gfc_get_dataptr_offset): Correcting call of build_array_ref.
(gfc_conv_expr_descriptor): Set the array's offset to -1 when
lbound in inner most dim is 1 and symbol non-pointer/assoc.
* trans-decl.c (gfc_build_qualified_array): Select correct
gfc_array_spec for regular and class arrays.
(gfc_build_dummy_array_decl): Likewise.
(gfc_get_symbol_decl): Get a dummy array for class arrays.
(gfc_trans_deferred_vars): Tell conv_expr that the descriptor
is desired.
* trans-expr.c (gfc_class_vptr_get): Get the class descriptor
from the correct location for class arrays.
(gfc_class_len_get): Likewise.
(gfc_conv_intrinsic_to_class): Add handling of _len component.
(gfc_conv_class_to_class):  Prevent access to unset array data
when the array is an optional argument. Add handling of _len
component.
(gfc_copy_class_to_class): Check that _def_init is non-NULL
when used in _vptr->copy()
(gfc_trans_class_init_assign): Ensure that the rank of
_def_init is zero.
(gfc_conv_component_ref): Get the _vptr along with _data refs.
(gfc_conv_variable): Make sure the temp array descriptor is
returned for class arrays, too, and that class arrays are
dereferenced correctly.
(gfc_conv_procedure_call): For polymorphic type initialization
the initializer has to be a pointer to _def_init stored in a
dummy variable, which then needs to be used by value.
* trans-intrinsic.c (gfc_conv_intrinsic_sizeof): Use the
temporary array descriptor for class arrays, too.
(gfc_conv_intrinsic_storage_size): Likewise.
(gfc_conv_intrinsic_loc): Add ref to _data for BT_CLASS
expressions.
* trans-stmt.c (trans_associate_var): Use a temporary array for
the associate variable of class arrays, too, making the array
one-based (lbound == 1).
* trans-types.c (gfc_is_nodesc_array): Use the correct
array data.
* trans.c (gfc_build_array_ref): Use the dummy array descriptor
when present.
* trans.h: Add class_vptr to gfc_se for storing a class ref's
vptr.

From-SVN: r222361

passes.def: Remove copy propagation passes run directly after CCP.

2015-04-23  Richard Biener  <rguenther@suse.de>

* passes.def: Remove copy propagation passes run directly after CCP.
* tree-ssa-ccp.c (get_value_for_expr): Fall back to a COPY for
SSA names.
(ccp_visit_phi_node): Rework to handle first executable edge
specially.

From-SVN: r222360

[ARM] Remove an unused reload hook.

* config/arm/arm.h (LEGITIMIZE_RELOAD_ADDRESS): Remove.
(ARM_LEGITIMIZE_RELOAD_ADDRESS): Remove.
(THUMB_LEGITIMIZE_RELOAD_ADDRESS): Remove.
* config/arm/arm.c (arm_legimitimize_reload_address): Remove.
(thumb_legimitimize_reload_address): Remove.
* config/arm/arm-protos.h (arm_legimitimize_reload_address):
Remove.
(thumb_legimitimize_reload_address): Remove.

From-SVN: r222359

Re: [PATCH 02/12] remove some ifdef HAVE_cc0.

* conditions.h (CC_STATUS_INIT): Gate on #ifndef CC_STATUS_INIT.

From-SVN: r222358

[ARM] Restrict {load,store}_multiple expanders to MAX_LD_STM_OPS regs

* config/arm/arm.md (load_multiple): Reject operand 2 greater than
MAX_LDM_STM_OPS.
(store_multiple): Likewise.

From-SVN: r222357

[ARM] Make issue rate part of per-core tuning structs

* config/arm/arm-protos.h (struct tune_params): Add issue_rate field.
* config/arm/arm.c (arm_slowmul_tune, arm_fastmul_tune,
arm_strongarm_tune, arm_xscale_tune, arm_9e_tune, arm_v6t2_tune,
arm_cortex_tune, arm_cortex_a8_tune, arm_cortex_a7_tune,
arm_cortex_a15_tune, arm_cortex_a53_tune, arm_cortex_a57_tune,
arm_cortex_a9_tune, arm_cortex_a12_tune, arm_v7m_tune, arm_v6m_tune,
arm_fa726te_tune arm_cortex_a5_tune, arm_xgene1_tune):
Specify issue_rate value.
(arm_issue_rate): Look up issue rate from tuning structs. Remove
large switch statement.
(arm_marvell_pj4_tune): New struct.
* config/arm/arm-cores.def (marvell-pj4): Use arm_marvell_pj4_tune
struct.

From-SVN: r222356

tree-vect-slp.c (vect_find_first_load_in_slp_instance): Remove.

2015-04-23  Richard Biener  <rguenther@suse.de>

* tree-vect-slp.c (vect_find_first_load_in_slp_instance): Remove.
(vect_find_last_store_in_slp_instance): Rename to ...
(vect_find_last_scalar_stmt_in_slp): ... this and generalize.
(vect_analyze_slp_cost_1): Use vector_load for constant defs
and vec_construct for external defs when estimating prologue cost.
(vect_analyze_slp_instance): Do not init SLP_INSTANCE_FIRST_LOAD_STMT.
Compute costs here only when vectorizing loops.
(vect_slp_analyze_bb_1): Compute SLP cost here, after vector types
have been determined.
(vect_schedule_slp_instance): Simplify vectorized code placement
and prepare for in-BB external defs.
* tree-vectorizer.h (struct _slp_instance): Remove first_load member.
(SLP_INSTANCE_FIRST_LOAD_STMT): Remove.
* tree-vect-stmts.c (vect_model_store_cost): Remove PURE_SLP_STMT
guard.
(vect_model_load_cost): Likewise.
(vectorizable_store): Instead add it here.
(vectorizable_load): Likewise.
(vect_is_simple_use): Dump def type textually.

From-SVN: r222354

cfgexpand.c (expand_gimple_stmt_1): Use ops.code.

2015-04-23  Richard Biener  <rguenther@suse.de>

* cfgexpand.c (expand_gimple_stmt_1): Use ops.code.
* cfgloop.c (verify_loop_structure): Verify the root loop node.
* except.c (duplicate_eh_regions): Call get_eh_region_from_lp_number_fn
instead of get_eh_region_from_lp_number.
* loop-init.c (fix_loop_structure): If we removed a loop, reset
the SCEV cache.

From-SVN: r222353

rs6000.c (rs6000_output_function_prologue): No need for -mprofile-kernel to save LR to stack.

* config/rs6000/rs6000.c (rs6000_output_function_prologue): No
need for -mprofile-kernel to save LR to stack.

From-SVN: r222352

rs6000.c (rtx_is_swappable_p): Commentary adjustments.

[gcc]

2015-04-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/rs6000.c (rtx_is_swappable_p): Commentary
adjustments.
(insn_is_swappable_p): Return 1 for a convert from double to
single precision when all of its uses are splats of BE element
zero.

[gcc/testsuite]

2015-04-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gcc.target/powerpc/swaps-p8-18.c: New test.

From-SVN: r222351

ira-costs.c (record_operand_costs): Fix typo (remove redundant code).

2015-04-23  Kugan Vivekanandarajah  <kuganv@linaro.org>

* ira-costs.c (record_operand_costs): Fix typo (remove redundant code).

From-SVN: r222350

re PR target/65456 (powerpc64le autovectorized copy loop missed optimization)

[gcc]

2015-04-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

PR target/65456
* config/rs6000/rs6000.c (rs6000_option_override_internal):  For
VSX + POWER8, enable TARGET_ALLOW_MOVMISALIGN and
TARGET_EFFICIENT_UNALIGNED_VSX if not selected by command line
option.
(rs6000_builtin_mask_for_load): Return 0 for targets with
efficient unaligned VSX accesses so that the vectorizer will use
direct unaligned loads.
(rs6000_builtin_support_vector_misalignment): Always return true
for targets with efficient unaligned VSX accesses.
(rs6000_builtin_vectorization_cost): Cost of unaligned loads and
stores on targets with efficient unaligned VSX accesses is almost
always the same as the cost of an aligned load or store, so model
it that way.
* config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Return 0 for
unaligned vectors if we have efficient unaligned VSX accesses.
* config/rs6000/rs6000.opt (mefficient-unaligned-vector): New
undocumented option.

[gcc/testsuite]

2015-04-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

PR target/65456
* gcc.dg/vect/bb-slp-24.c: Exclude test for POWER8.
* gcc.dg/vect/bb-slp-25.c: Likewise.
* gcc.dg/vect/bb-slp-29.c: Likewise.
* gcc.dg/vect/bb-slp-32.c: Replace vect_no_align with
vect_no_align && { ! vect_hw_misalign }.
* gcc.dg/vect/bb-slp-9.c: Likewise.
* gcc.dg/vect/costmodel/ppc/costmodel-slp-33.c: Exclude test for
vect_hw_misalign.
* gcc.dg/vect/costmodel/ppc/costmodel-vect-31a.c: Likewise.
* gcc.dg/vect/costmodel/ppc/costmodel-vect-76b.c: Adjust tests to
account for POWER8, where peeling for alignment is not needed.
* gcc.dg/vect/costmodel/ppc/costmodel-vect-outer-fir.c: Replace
vect_no_align with vect_no_align && { ! vect_hw_misalign }.
* gcc.dg.vect.if-cvt-stores-vect-ifcvt-18.c: Likewise.
* gcc.dg/vect/no-scevccp-outer-6-global.c: Likewise.
* gcc.dg/vect/no-scevccp-outer-6.c: Likewise.
* gcc.dg/vect/no-vfa-vect-43.c: Likewise.
* gcc.dg/vect/no-vfa-vect-57.c: Likewise.
* gcc.dg/vect/no-vfa-vect-61.c: Likewise.
* gcc.dg/vect/no-vfa-vect-depend-1.c: Likewise.
* gcc.dg/vect/no-vfa-vect-depend-2.c: Likewise.
* gcc.dg/vect/no-vfa-vect-depend-3.c: Likewise.
* gcc.dg/vect/pr16105.c: Likewise.
* gcc.dg/vect/pr20122.c: Likewise.
* gcc.dg/vect/pr33804.c: Likewise.
* gcc.dg/vect/pr33953.c: Likewise.
* gcc.dg/vect/pr56787.c: Likewise.
* gcc.dg/vect/pr58508.c: Likewise.
* gcc.dg/vect/slp-25.c: Likewise.
* gcc.dg/vect/vect-105-bit-array.c: Likewise.
* gcc.dg/vect/vect-105.c: Likewise.
* gcc.dg/vect/vect-27.c: Likewise.
* gcc.dg/vect/vect-29.c: Likewise.
* gcc.dg/vect/vect-33.c: Exclude unaligned access test for
POWER8.
* gcc.dg/vect/vect-42.c: Replace vect_no_align with vect_no_align
&& { ! vect_hw_misalign }.
* gcc.dg/vect/vect-44.c: Likewise.
* gcc.dg/vect/vect-48.c: Likewise.
* gcc.dg/vect/vect-50.c: Likewise.
* gcc.dg/vect/vect-52.c: Likewise.
* gcc.dg/vect/vect-56.c: Likewise.
* gcc.dg/vect/vect-60.c: Likewise.
* gcc.dg/vect/vect-72.c: Likewise.
* gcc.dg/vect/vect-75-big-array.c: Likewise.
* gcc.dg/vect/vect-75.c: Likewise.
* gcc.dg/vect/vect-77-alignchecks.c: Likewise.
* gcc.dg/vect/vect-77-global.c: Likewise.
* gcc.dg/vect/vect-78-alignchecks.c: Likewise.
* gcc.dg/vect/vect-78-global.c: Likewise.
* gcc.dg/vect/vect-93.c: Likewise.
* gcc.dg/vect/vect-95.c: Likewise.
* gcc.dg/vect/vect-96.c: Likewise.
* gcc.dg/vect/vect-cond-1.c: Likewise.
* gcc.dg/vect/vect-cond-3.c: Likewise.
* gcc.dg/vect/vect-cond-4.c: Likewise.
* gcc.dg/vect/vect-cselim-1.c: Likewise.
* gcc.dg/vect/vect-multitypes-1.c: Likewise.
* gcc.dg/vect/vect-multitypes-3.c: Likewise.
* gcc.dg/vect/vect-multitypes-4.c: Likewise.
* gcc.dg/vect/vect-multitypes-6.c: Likewise.
* gcc.dg/vect/vect-nest-cycle-1.c: Likewise.
* gcc.dg/vect/vect-nest-cycle-2.c: Likewise.
* gcc.dg/vect/vect-outer-3a-big-array.c: Likewise.
* gcc.dg/vect/vect-outer-3a.c: Likewise.
* gcc.dg/vect/vect-outer-5.c: Likewise.
* gcc.dg/vect/vect-outer-fir-big-array.c: Likewise.
* gcc.dg/vect/vect-outer-fir-lb-big-array.c: Likewise.
* gcc.dg/vect/vect-outer-fir-lb.c: Likewise.
* gcc.dg/vect/vect-outer-fir.c: Likewise.
* gcc.dg/vect/vect-peel-3.c: Likewise.
* gcc.dg/vect/vect-peel-4.c: Likewise.
* gcc.dg/vect/vect-pre-interact.c: Likewise.
* gcc.target/powerpc/pr65456.c: New test.
* gcc.target/powerpc/vsx-vectorize-2.c: Exclude test for POWER8.
* gcc.target/powerpc/vsx-vectorize-4.c: Likewise.
* gcc.target/powerpc/vsx-vectorize-6.c: Likewise.
* gcc.target/powerpc/vsx-vectorize-7.c: Likewise.
* gfortran.dg/vect/vect-2.f90: Replace vect_no_align with
vect_no_align && { ! vect_hw_misalign }.
* gfortran.dg/vect/vect-3.f90: Likewise.
* gfortran.dg/vect/vect-4.f90: Likewise.
* gfortran.dg/vect/vect-5.f90: Likewise.
* lib/target-supports.exp (check_effective_target_vect_no_align):
Return 1 for POWER8.
(check_effective_target_vect_hw_misalign): Return 1 for POWER8.

From-SVN: r222349

Daily bump.

From-SVN: r222348

re PR fortran/65429 (ICE on implied-length character empty array constructor)

2015-04-22  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/65429
* decl.c (add_init_expr_to_sym): Set the length type parameter.

PR fortran/65429
* gfortran.dg/pr65429.f90: New test.

From-SVN: r222342

libiberty/setenv.c: Do not declare environ if defined as a macro.

2015-04-22  Eli Zaretskii  <eliz@gnu.org>

        * strerror.c <sys_nerr, sys_errlist>: Declare only if they aren't
        macros.
        * setenv.c <environ>: Declare only if not a macro.

From-SVN: r222335

Makefile.tpl: Remove surplus whitespace throughout.

2015-04-22  Bernhard Reutner-Fischer  <aldot@gcc.gnu.org>

* Makefile.tpl: Remove surplus whitespace throughout.
* Makefile.in: Regenerate.

From-SVN: r222334

Revert:

2015-04-22  Gregor Richards  <gregor.richards@uwaterloo.ca>

* config.gcc (LIBC_MUSL): New tm_defines macro.
* config/linux.h (OPTION_MUSL): Define.
(INCLUDE_DEFAULTS_MUSL_GPP, INCLUDE_DEFAULTS_MUSL_LOCAL,)
(INCLUDE_DEFAULTS_MUSL_PREFIX, INCLUDE_DEFAULTS_MUSL_CROSS,)
(INCLUDE_DEFAULTS_MUSL_TOOL, INCLUDE_DEFAULTS_MUSL_NATIVE): Define.

* config/linux.opt (mmusl): New option.
* configure.ac (gcc_cv_libc_provides_ssp): Add *-*-musl*.
(gcc_cv_target_dl_iterate_phdr): Add *-linux-musl*.

* configure: Regenerate.

From-SVN: r222330

libstdc++, libgfortran gthr workaround for musl

On behalf of szabolcs.nagy@arm.com

[libstdc++-v3/]
2015-04-22  Szabolcs Nagy  <szabolcs.nagy@arm.com>

* config/os/generic/os_defines.h (_GLIBCXX_GTHREAD_USE_WEAK): Define.
* configure.host (os_include_dir): Set to "os/generic" for linux-musl*.

[libgfortran/]
2015-04-22  Szabolcs Nagy  <szabolcs.nagy@arm.com>

* acinclude.m4 (GTHREAD_USE_WEAK): Define as 0 for *-*-musl*.
* configure: Regenerate.

From-SVN: r222329

unwind fix for musl

On behalf of szabolcs.nagy@arm.com

2015-04-22  Gregor Richards  <gregor.richards@uwaterloo.ca>
    Szabolcs Nagy  <szabolcs.nagy@arm.com>

* unwind-dw2-fde-dip.c (USE_PT_GNU_EH_FRAME): Define it on
Linux if target provides dl_iterate_phdr.

Co-Authored-By: Szabolcs Nagy <szabolcs.nagy@arm.com>
From-SVN: r222328

fixincludes update for musl support

On behalf of Szabolcs.Nagy@arm.com

2015-04-22  Gregor Richards  <gregor.richards@uwaterloo.ca>

* mkfixinc.sh: Add *-musl* with no fixes.

From-SVN: r222327

config changes for musl libc support

On behalf of Szabolcs.Nagy@arm.com

2015-04-22  Gregor Richards  <gregor.richards@uwaterloo.ca>

* config.gcc (LIBC_MUSL): New tm_defines macro.
* config/linux.h (OPTION_MUSL): Define.
(INCLUDE_DEFAULTS_MUSL_GPP, INCLUDE_DEFAULTS_MUSL_LOCAL,)
(INCLUDE_DEFAULTS_MUSL_PREFIX, INCLUDE_DEFAULTS_MUSL_CROSS,)
(INCLUDE_DEFAULTS_MUSL_TOOL, INCLUDE_DEFAULTS_MUSL_NATIVE): Define.

* config/linux.opt (mmusl): New option.
* configure.ac (gcc_cv_libc_provides_ssp): Add *-*-musl*.
(gcc_cv_target_dl_iterate_phdr): Add *-linux-musl*.

* configure: Regenerate.

From-SVN: r222326

libitm fixes for musl support

On behalf of Szabolcs.Nagy@arm.com

2015-04-22  Gregor Richards  <gregor.richards@uwaterloo.ca>

       * config/arm/hwcap.cc: Use fcntl.h instead of sys/fcntl.h.
       * config/linux/x86/tls.h: Only use __GLIBC_PREREQ if defined.

From-SVN: r222325

[PATCH][libstc++v3]Add new dg-require-thread-fence directive.

libstdc++-v3/

2015-04-22  Renlin Li  <renlin.li@arm.com>

* testsuite/lib/dg-options.exp (dg-require-thread-fence): New.
* testsuite/lib/libstdc++.exp (check_v3_target_thread_fence): New.
* testsuite/29_atomics/atomic_flag/clear/1.cc: Use it.
* testsuite/29_atomics/atomic_flag/test_and_set/explicit.cc: Likewise.
* testsuite/29_atomics/atomic_flag/test_and_set/implicit.cc: Likewise.

From-SVN: r222324

Allow wildcards in -fsanitize-sections.

2015-04-22  Yury Gribov  <y.gribov@samsung.com>

Allow wildcards in -fsanitize-sections.

gcc/
* asan.c (set_sanitized_sections): Parse incoming arg.
(section_sanitized_p): Support wildcards.
* doc/invoke.texi (-fsanitize-sections): Update description.

gcc/testsuite/
* c-c++-common/asan/user-section-1.c: New test.
* c-c++-common/asan/user-section-2.c: New test.
* c-c++-common/asan/user-section-3.c: New test.

From-SVN: r222322

Fix va_arg ap_copy nop detection

2015-04-22  Tom de Vries  <tom@codesourcery.com>

PR tree-optimization/65823
* gimplify.c (gimplify_modify_expr): Use operand_equal_p to test for
equality between ap_copy and ap.

From-SVN: r222318

PR target/47098 OBSD_LIB_SPEC on i686-openbsd3 missing

config-list.mk for i686-*-openbsd3.0 failed with:

gcc/config/openbsd.h:143:18: error: ‘OBSD_LIB_SPEC’ was not declared in this scope
 #define LIB_SPEC OBSD_LIB_SPEC
                  ^
gcc/gcc.c:879:31: note:
in expansion of macro ‘LIB_SPEC’
 static const char *lib_spec = LIB_SPEC;
                               ^
make[2]: *** [gcc.o] Error 1

i.e. openbsd-oldgas.h is only used on
i?86-*-openbsd2.*|i?86-*openbsd3.[0123] so add the OBSD_LIB_SPEC there

From-SVN: r222315

PR target/47122: vax-*-openbsd* config.gcc typo

Gerald / jsg typoed this in the initial r152218 a.k.a
8e6600a15374f9446d5026e8ef6b68e0d753fb6c

From-SVN: r222314

re PR target/55144 (opening glibc-c.o: No such file or directory)

PR target/55144

building all-gcc for bfin-linux-uclibc resulted in

build/genchecksum cp/cp-lang.o c-family/stub-objc.o ... glibc-c.o \
libbackend.a .. > cc1plus-checksum.c.tmp
opening glibc-c.o: No such file or directory
make[2]: *** [cc1-checksum.c] Error 1

From-SVN: r222313

tree-tailcall.c (suitable_for_tail_opt_p, [...]): Remove unneeded forward declarations.

2015-04-22  Bernhard Reutner-Fischer  <aldot@gcc.gnu.org>

* tree-tailcall.c (suitable_for_tail_opt_p, find_tail_calls):
Remove unneeded forward declarations.
(suitable_for_tail_call_opt_p): Commentary typo fix.

From-SVN: r222312

varasm.c (emit_bss): Remove redundant guard.

The whole emit_bss is wrapped in ifdef ASM_OUTPUT_ALIGNED_BSS

From-SVN: r222311

* gennews (files): Add files for GCC 5.

From-SVN: r222309

c6x: handle c6x unk_isa in TARGET_CPU_CPP_BUILTINS

Fixes all-gcc for c6x config-list.mk that complained about

error: enumeration value ‘unk_isa’ not handled in switch

Maybe seeing unk_isa in the cpu switch statement should
__builtin_abort() but be conservative to just allow all-gcc of config-list.mk
to compile

From-SVN: r222308

bfin: handle BFIN_CPU_UNKNOWN in TARGET_CPU_CPP_BUILTINS

gcc/c-family/c-cppbuiltin.c
In file included from ./tm.h:21:0,
                 from
../../../../../../home/me/src/gcc-5.0.mine/gcc/c-family/c-cppbuiltin.c:23:
../../../../../../home/me/src/gcc-5.0.mine/gcc/c-family/c-cppbuiltin.c:
In function ‘void c_cpp_builtins(cpp_reader*)’:
../../../../../../home/me/src/gcc-5.0.mine/gcc/config/bfin/bfin.h:43:14:
error: enumeration value ‘BFIN_CPU_UNKNOWN’ not handled in switch
[-Werror=switch]
       switch (bfin_cpu_type)   \
              ^
../../../../../../home/me/src/gcc-5.0.mine/gcc/c-family/c-cppbuiltin.c:1243:3:
note: in expansion of macro ‘TARGET_CPU_CPP_BUILTINS’
   TARGET_CPU_CPP_BUILTINS ();
   ^
cc1plus: all warnings being treated as errors
make[2]: *** [c-family/c-cppbuiltin.o] Error 1

Maybe seeing BFIN_CPU_UNKNOWN in the cpu switch statement should
__builtin_abort() but be conservative to just allow all-gcc of config-list.mk
to compile

From-SVN: r222307

re PR rtl-optimization/64818 (User specified register don't work correctly in inline-asm operands.)

gcc/ChangeLog:
2015-04-22  Hale Wang  <hale.wang@arm.com>
            Terry Guo  <terry.guo@arm.com>

       PR rtl-optimization/64818
       * combine.c (can_combine_p): Don't combine user-specified
       register if it is in an asm input.

gcc/testsuite/ChangeLog
2015-04-22  Hale Wang  <hale.wang@arm.com>
            Terry Guo  <terry.guo@arm.com>

       PR rtl-optimization/64818
       * gcc.target/arm/pr64818.c: New test.

Co-Authored-By: Terry Guo <terry.guo@arm.com>
From-SVN: r222306

re PR ipa/65076 (16% tramp3d-v4.cpp compile time regression)

PR ipa/65076
* passes.def (early_optimizations): Add pass_dse.

* g++.dg/tree-ssa/pr61034.C: Update template.
* g++.dg/warn/Warray-bounds.C: Harden for DSE.
* gcc.dg/Warray-bounds-11.c: Likewise.
* gcc.dg/Warray-bounds.c: Likewise.

From-SVN: r222305

add default for INSN_REFERENCES_ARE_DELAYED

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* defaults.h (INSN_REFERENCES_ARE_DELAYED): New definition.
* reorg.c (redundant_insn): Remove ifdef
INSN_REFERENCES_ARE_DELAYED.
* resource.c (mark_referenced_resources): Likewise.

From-SVN: r222304

provide default for INSN_SETS_ARE_DELAYED

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* defaults.h (INSN_SETS_ARE_DELAYED): New definition.
* reorg.c (redundant_insn): Remove ifdef INSN_SETS_ARE_DELAYED.
* resource.c (mark_set_resources): Likewise.

From-SVN: r222303

remove more ifdefs for HAVE_cc0

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* caller-save.c (insert_one_insn): Remove ifdef HAVE_cc0.
* cfgcleanup.c (flow_find_cross_jump): Likewise.
(flow_find_head_matching_sequence): Likewise.
(try_head_merge_bb): Likewise.
* combine.c (can_combine_p): Likewise.
(try_combine): Likewise.
(distribute_notes): Likewise.
* df-problems.c (can_move_insns_across): Likewise.
* final.c (final): Likewise.
* gcse.c (insert_insn_end_basic_block): Likewise.
* ira.c (find_moveable_pseudos): Likewise.
* reorg.c (try_merge_delay_insns): Likewise.
(fill_simple_delay_slots): Likewise.
(fill_slots_from_thread): Likewise.
* sched-deps.c (sched_analyze_2): Likewise.

From-SVN: r222302

remove #if for PIC_OFFSET_TABLE_REGNUM

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* df-scan.c (df_get_entry_block_def_set): Remove #ifdef
PIC_OFFSET_TABLE_REGNUM.

From-SVN: r222301

reduce conditional compilation for HARD_FRAME_POINTER_IS_FRAME_POINTER

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* alias.c (init_alias_target): Remove ifdef
* HARD_FRAME_POINTER_IS_FRAME_POINTER.
* df-scan.c (df_insn_refs_collect): Likewise.
(df_get_regular_block_artificial_uses): Likewise.
(df_get_eh_block_artificial_uses): Likewise.
(df_get_entry_block_def_set): Likewise.
(df_get_exit_block_use_set): Likewise.
* emit-rtl.c (gen_rtx_REG): Likewise.
* ira.c (ira_setup_eliminable_regset): Likewise.
* reginfo.c (init_reg_sets_1): Likewise.
* regrename.c (rename_chains): Likewise.
* reload1.c (reload): Likewise.
(eliminate_regs_in_insn): Likewise.
* resource.c (mark_referenced_resources): Likewise.
(init_resource_info): Likewise.

From-SVN: r222300

provide default for MASK_RETURN_ADDR

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* defaults.h (MASK_RETURN_ADDR): New definition.
* except.c (expand_builtin_extract_return_addr): Remove ifdef
MASK_RETURN_ADDR.

From-SVN: r222299

provide default for RETURN_ADDR_OFFSET

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* defaults.h (RETURN_ADDR_OFFSET): New definition.
* except.c (expand_builtin_extract_return_addr): Remove ifdef
RETURN_ADDR_OFFSET.
(expand_builtin_frob_return_addr): Likewise.

From-SVN: r222298

make some HAVE_cc0 code always compiled

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* cfgrtl.c (rtl_merge_blocks): Change #if HAVE_cc0 to if (HAVE_cc0)
(try_redirect_by_replacing_jump): Likewise.
(rtl_tidy_fallthru_edge): Likewise.
* combine.c (insn_a_feeds_b): Likewise.
(find_split_point): Likewise.
(simplify_set): Likewise.
* cprop.c (cprop_jump): Likewise.
* cse.c (cse_extended_basic_block): Likewise.
* df-problems.c (can_move_insns_across): Likewise.
* function.c (emit_use_return_register_into_block): Likewise.
* haifa-sched.c (sched_init): Likewise.
* ira.c (find_moveable_pseudos): Likewise.
* loop-invariant.c (find_invariant_insn): Likewise.
* lra-constraints.c (curr_insn_transform): Likewise.
* postreload.c (reload_combine_recognize_const_pattern):
* Likewise.
* reload.c (find_reloads): Likewise.
* reorg.c (delete_scheduled_jump): Likewise.
(steal_delay_list_from_target): Likewise.
(steal_delay_list_from_fallthrough): Likewise.
(redundant_insn): Likewise.
(fill_simple_delay_slots): Likewise.
(fill_slots_from_thread): Likewise.
(delete_computation): Likewise.
* sched-rgn.c (add_branch_dependences): Likewise.

From-SVN: r222297

always define HAVE_cc0

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* genconfig.c (main): Always define HAVE_cc0.
* caller-save.c (insert_one_insn): Change ifdef HAVE_cc0 to #if
HAVE_cc0.
* cfgcleanup.c (flow_find_cross_jump): Likewise.
(flow_find_head_matching_sequence): Likewise.
(try_head_merge_bb): Likewise.
* cfgrtl.c (rtl_merge_blocks): Likewise.
(try_redirect_by_replacing_jump): Likewise.
(rtl_tidy_fallthru_edge): Likewise.
* combine.c (do_SUBST_MODE): Likewise.
(insn_a_feeds_b): Likewise.
(combine_instructions): Likewise.
(can_combine_p): Likewise.
(try_combine): Likewise.
(find_split_point): Likewise.
(subst): Likewise.
(simplify_set): Likewise.
(distribute_notes): Likewise.
* cprop.c (cprop_jump): Likewise.
* cse.c (cse_extended_basic_block): Likewise.
* df-problems.c (can_move_insns_across): Likewise.
* final.c (final): Likewise.
(final_scan_insn): Likewise.
* function.c (emit_use_return_register_into_block): Likewise.
* gcse.c (insert_insn_end_basic_block): Likewise.
* haifa-sched.c (sched_init): Likewise.
* ira.c (find_moveable_pseudos): Likewise.
* loop-invariant.c (find_invariant_insn): Likewise.
* lra-constraints.c (curr_insn_transform): Likewise.
* optabs.c (prepare_cmp_insn): Likewise.
* postreload.c (reload_combine_recognize_const_pattern):
* Likewise.
* reload.c (find_reloads): Likewise.
(find_reloads_address_1): Likewise.
* reorg.c (delete_scheduled_jump): Likewise.
(steal_delay_list_from_target): Likewise.
(steal_delay_list_from_fallthrough): Likewise.
(try_merge_delay_insns): Likewise.
(redundant_insn): Likewise.
(fill_simple_delay_slots): Likewise.
(fill_slots_from_thread): Likewise.
(delete_computation): Likewise.
(relax_delay_slots): Likewise.
* sched-deps.c (sched_analyze_2): Likewise.
* sched-rgn.c (add_branch_dependences): Likewise.

From-SVN: r222296

remove more ifdef HAVE_cc0

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* combine.c (find_single_use): Remove HAVE_cc0 ifdef for code
that is trivially ded on non cc0 targets.
(simplify_set): Likewise.
(mark_used_regs_combine): Likewise.
* cse.c (new_basic_block): Likewise.
(fold_rtx): Likewise.
(cse_insn): Likewise.
(cse_extended_basic_block): Likewise.
(set_live_p): Likewise.
* rtlanal.c (canonicalize_condition): Likewise.
* simplify-rtx.c (simplify_binary_operation_1): Likewise.

From-SVN: r222295

remove some ifdef HAVE_cc0

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* conditions.h: Define macros even if HAVE_cc0 is undefined.
* emit-rtl.c: Define functions even if HAVE_cc0 is undefined.
* final.c: Likewise.
* jump.c: Likewise.
* recog.c: Likewise.
* recog.h: Declare functions even when HAVE_cc0 is undefined.
* sched-deps.c (sched_analyze_2): Always compile case for cc0.

From-SVN: r222294

add default definition of EH_RETURN_DATA_REGNO

gcc/ChangeLog:

2015-04-21  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* defaults.h: New definition of EH_RETURN_DATA_REGNO.
* except.c: Remove definition of EH_RETURN_DATA_REGNO.
* builtins.c (expand_builtin): Remove check if
EH_RETURN_DATA_REGNO is defined.
* df-scan.c (df_bb_refs_collect): Likewise.
(df_get_exit_block_use_set): Likewise.
* haifa-sched.c (initiate_bb_reg_pressure_info): Likewise.
* ira-lives.c (process_bb_node_lives): Likewise.
* lra-lives.c (process_bb_lives): Likewise.

From-SVN: r222293

Daily bump.

From-SVN: r222292

re PR libfortran/65234 (Output descriptor (*(1E15.7)) not accepted)

2015-04-21 Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/65234
* gfortran.dg/fmt_unlimited.f90: New test.

From-SVN: r222276

re PR libfortran/65234 (Output descriptor (*(1E15.7)) not accepted)

2015-04-21 Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/65234
* io/format.c (parse_format_list): Set the seen_dd flag in all
cases where a data descriptor has been seen.

From-SVN: r222274

avx512bw-vpermi2w-2.c: Fix includes to use actual headers.

    * gcc.target/i386/avx512bw-vpermi2w-2.c: Fix includes to use actual
    headers.
    * gcc.target/i386/avx512bw-vpermt2w-2.c: Likewise.
    * gcc.target/i386/avx512bw-vpmaddubsw-2.c: Likewise.
    * gcc.target/i386/avx512bw-vpmaddwd-2.c: Likewise.
    * gcc.target/i386/avx512dq-vfpclasspd-2.c: Likewise.
    * gcc.target/i386/avx512dq-vfpclassps-2.c: Likewise.
    * gcc.target/i386/avx512vbmi-vpermi2b-2.c: Likewise.
    * gcc.target/i386/avx512vbmi-vpermt2b-2.c: Likewise.

From-SVN: r222273

re PR fortran/56743 (Namelist bug with comment and no blank)

2015-04-21 Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/56743
* gfortran.dg/namelist_87.f90: New test.

From-SVN: r222272

re PR fortran/56743 (Namelist bug with comment and no blank)

2015-04-21 Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/56743
* io/list_read.c (CASE_SEPARATORS): Add case for '!'.
(is_separator): Add condition for '!'.
(eat_separator): Use notify_std to warn or errord if '!' is
encountered before a proper separator.

From-SVN: r222271

i386.md (ARGP_REG, [...]): New.

* config/i386/i386.md (ARGP_REG, FRAME_REG, BND2_REG, BND3_REG,
FIRST_PSEUDO_REG): New.
* config/i386/i386.h (STACK_POINTER_REGNUM): Define to SP_REG.
(ARG_POINTER_REGNUM): Define to ARGP_REG.
(FRAME_POINTER_REGNUM): Define to FRAME_REG.
(HARD_FRAME_POINTER_REGNUM): Define to BP_REG.
(FIRST_PSEUDO_REGISTER): Define to FIRST_PSEUDO_REG.
(FIRST_INT_REG): New.
(LAST_INT_REG): New.
(FIRST_*_REG): Define using *_REG.
(LAST_*_REG): Ditto.
(QI_REGNO_P): Define using FIRST_QU_REG and LAST_QI_REG.
(LEGACY_INT_REGNO_P): Define using FIRST_INT_REG and LAST_INT_REG.
(FIRST_FLOAT_REG): Define to FIRST_STACK_REG.

From-SVN: r222269

[expmed] Properly account for the cost and latency of shift+add ops when synthesizing mults

        * expmed.c: (synth_mult): Only assume overlapping
        shift with previous steps in alg_sub_t_m2 case.

        * gcc.target/aarch64/mult-synth_1.c: New test.
        * gcc.target/aarch64/mult-synth_2.c: Likewise.
        * gcc.target/aarch64/mult-synth_3.c: Likewise.
        * gcc.target/aarch64/mult-synth_4.c: Likewise.
        * gcc.target/aarch64/mult-synth_5.c: Likewise.
        * gcc.target/aarch64/mult-synth_6.c: Likewise.

From-SVN: r222268

re PR tree-optimization/65650 (CCP does not propgate copies)

2015-04-21  Richard Biener  <rguenther@suse.de>

PR tree-optimization/65650
* tree-ssa-ccp.c (valid_lattice_transition): Allow lattice
transitions involving copies.
(set_lattice_value): Adjust for copy lattice state.
(ccp_lattice_meet): Do not merge UNDEFINED and a copy to the copy
if that doesn't dominate the merge point.
(bit_value_unop): Adjust what we treat as varying mask.
(bit_value_binop): Likewise.
(bit_value_assume_aligned): Likewise.
(evaluate_stmt): When we simplified to a SSA name record a copy
instead of dropping to varying.
(visit_assignment): Simplify.

* gimple-match.h (gimple_simplify): Add another callback.
* gimple-fold.c (fold_stmt_1): Adjust caller.
(gimple_fold_stmt_to_constant_1): Likewise - pass valueize
for the 2nd callback.
* gimple-match-head.c (gimple_simplify): Add a callback that is
used to valueize the stmt operands and use it that way.

* gcc.dg/tree-ssa/ssa-ccp-37.c: New testcase.
* gcc.dg/tree-ssa/forwprop-11.c: Adjust.
* gcc.dg/tree-ssa/ssa-fre-3.c: Likewise.
* gcc.dg/tree-ssa/ssa-fre-4.c: Likewise.
* gcc.dg/tree-ssa/ssa-fre-5.c: Likewise.
* gcc.dg/tree-ssa/ssa-fre-32.c: Likewise.

From-SVN: r222267

re PR middle-end/65788 (416.gamess in SPEC CPU 2006 failed to build)

2015-04-21  Richard Biener  <rguenther@suse.de>

PR tree-optimization/65788
* tree-ssa-ccp.c (evaluate_stmt): Evaluate to UNDEFINED early.

From-SVN: r222266

i386.c (ix86_builtin_vectorization_cost): Scale vec_construct cost by vec_stmt_cost.

2015-04-21  Richard Biener  <rguenther@suse.de>

* config/i386/i386.c (ix86_builtin_vectorization_cost): Scale
vec_construct cost by vec_stmt_cost.

From-SVN: r222265

cfghooks.h (create_basic_block): Replace with two overloads for RTL and GIMPLE.

2015-04-21  Richard Biener  <rguenther@suse.de>

* cfghooks.h (create_basic_block): Replace with two overloads
for RTL and GIMPLE.
(split_block): Likewise.
* cfghooks.c (split_block): Rename to ...
(split_block_1): ... this.
(split_block): Add two type-safe overloads for RTL and GIMPLE.
(split_block_after_labels): Call split_block_1.
(create_basic_block): Rename to ...
(create_basic_block_1): ... this.
(create_basic_block): Add two type-safe overloads for RTL and GIMPLE.
(create_empty_bb): Call create_basic_block_1.
* cfgrtl.c (fixup_fallthru_exit_predecessor): Use
split_block_after_labels.
* omp-low.c (expand_parallel_call): Likewise.
(expand_omp_target): Likewise.
(simd_clone_adjust): Likewise.
* tree-chkp.c (chkp_get_entry_block): Likewise.
* cgraphunit.c (init_lowered_empty_function): Use the GIMPLE
create_basic_block overload.
(cgraph_node::expand_thunk): Likewise.
* tree-cfg.c (make_blocks): Likewise.
(handle_abnormal_edges): Likewise.
* tree-inline.c (copy_bb): Likewise.

From-SVN: r222264

[AArch64] Add zero_extend variants of logical+not ops

* config/aarch64/aarch64.md (*<NLOGICAL:optab>_one_cmplsidi3_ze):
New pattern.
(*xor_one_cmplsidi3_ze): Likewise.

From-SVN: r222263