re PR target/60697 ([aarch64] LRA ICE (Segfault) while building 435.gromacs)

2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>

PR target/60697
* lra-constraints.c (index_part_to_reg): New.
(process_address): Use it.

2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>

PR target/60697
* gcc.target/aarch64/pr60697.c: New.

From-SVN: r208926

Fixup ChangeLog formatting.

From-SVN: r208925

re PR target/60648 (ICE (segmentation fault) in expand_binop)

PR target/60648
       * expr.c (do_tablejump): Use simplify_gen_binary rather than
       gen_rtx_{PLUS,MULT} to build up the address expression.

       * i386/i386.c (ix86_legitimize_address): Use copy_addr_to_reg to avoid
       creating non-canonical RTL.

       PR target/60648
       * g++.dg/pr60648.C: New test.

Co-Authored-By: Jakub Jelinek <jakub@redhat.com>
From-SVN: r208924

re PR c++/60573 ([c++1y] ICE with defining generic function of nested class in class scope)

Fix PR c++/60573

PR c++/60573
* name-lookup.h (cp_binding_level): New transient field defining_class_p
to indicate whether a scope is in the process of defining a class.
* semantics.c (begin_class_definition): Set defining_class_p.
* name-lookup.c (leave_scope): Reset defining_class_p.
* parser.c (synthesize_implicit_template_parm): Use cp_binding_level::
defining_class_p rather than TYPE_BEING_DEFINED as the predicate for
unwinding to class-defining scope to handle the erroneous definition of
a generic function of an arbitrarily nested class within an enclosing
class.

PR c++/60573
* g++.dg/cpp1y/pr60573.C: New testcase.

From-SVN: r208921

Add PR60576 to the ChangeLog entry.

From-SVN: r208920

re PR fortran/60576 (FAIL: gfortran.dg/assumed_rank_7.f90)

2014-03-28  Mikael Morin  <mikael@gcc.gnu.org>
            Tobias Burnus  <burnus@net-b.de>

        PR fortran/60576
        * trans-expr.c (gfc_conv_derived_to_class): Avoid
        generation of out-of-bounds range expr.

Co-Authored-By: Tobias Burnus <burnus@net-b.de>
From-SVN: r208918

re PR ipa/60243 (IPA is slow on large cgraph tree)

PR ipa/60243
* ipa-inline.c (want_inline_small_function_p): Short circuit large
functions; reorganize to make cheap checks first.
(inline_small_functions): Do not estimate growth when dumping;
it is expensive.
* ipa-inline.h (inline_summary): Add min_size.
(growth_likely_positive): New function.
* ipa-inline-analysis.c (dump_inline_summary): Add min_size.
(set_cond_stmt_execution_predicate): Cleanup.
(estimate_edge_size_and_time): Compute min_size.
(estimate_calls_size_and_time): Likewise.
(estimate_node_size_and_time): Likewise.
(inline_update_overall_summary): Update min_size.
(do_estimate_edge_time): Likewise.
(do_estimate_edge_size): Update.
(do_estimate_edge_hints): Update.
(growth_likely_positive): New function.

From-SVN: r208916

re PR target/60693 (ICE on funny memcpy)

PR target/60693
* config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
also if addr has VOIDmode.

* gcc.target/i386/pr60693.c: New test.

From-SVN: r208915

pr59940.c (si): Use 32-bit SI mode instead of int.

2014-03-28  Vishnu K S  <Vishnu.k_s@atmel.com>

* gcc.dg/pr59940.c (si): Use 32-bit SI mode instead of int.

From-SVN: r208914

re PR fortran/60677 (FAIL: gfortran.dg/ichar_3.f90  -O  (test for excess errors))

fortran/
        PR fortran/60677
        * trans-intrinsic.c (gfc_conv_intrinsic_ichar): Enlarge argument
        list buffer.

From-SVN: r208913

re PR c++/60689 (Bogus error with atomic::exchange)

PR c++/60689
* c-tree.h (c_build_function_call_vec): New prototype.
* c-typeck.c (build_function_call_vec): Don't call
resolve_overloaded_builtin here.
(c_build_function_call_vec): New wrapper function around
build_function_call_vec.  Call resolve_overloaded_builtin here.
(convert_lvalue_to_rvalue, build_function_call, build_atomic_assign):
Call c_build_function_call_vec instead of build_function_call_vec.
* c-parser.c (c_parser_postfix_expression_after_primary): Likewise.
* c-decl.c (finish_decl): Likewise.

* c-common.c (add_atomic_size_parameter): When creating new
params vector, push the size argument first.

* c-c++-common/pr60689.c: New test.

From-SVN: r208912

[ARM/AArch64][2/2] Crypto intrinsics tuning for Cortex-A53 - pipeline description

* config/arm/aarch-common.c (aarch_crypto_can_dual_issue): New.
* config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
Declare extern.
* config/arm/cortex-a53.md: Add reservations and bypass for crypto
instructions as well as AdvancedSIMD loads.

From-SVN: r208910

[ARM/AArch64][1/2] Crypto intrinsics tuning for Cortex-A53 - "type" Attribute restructuring

     * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
     Use crypto_aese type.
     (aarch64_crypto_aes<aesmc_op>v16qi): Use crypto_aesmc type.
     * config/arm/arm.md (is_neon_type): Replace crypto_aes with
     crypto_aese, crypto_aesmc.  Move to types.md.
     * config/arm/types.md (crypto_aes): Split into crypto_aese,
     crypto_aesmc.
     * config/arm/iterators.md (crypto_type): Likewise.

From-SVN: r208908

re PR c++/58678 (pykde4-4.11.2 link error (devirtualization too trigger happy))

PR c++/58678
* g++.dg/abi/thunk6.C: Scan assembler for _ZTv0_n32_N1CD1Ev
only for lp64 targets and scan for _ZTv0_n16_N1CD1Ev for ilp32
targets.

From-SVN: r208907

cgraph.c: Include expr.h and tree-dfa.h.

* cgraph.c: Include expr.h and tree-dfa.h.
(cgraph_redirect_edge_call_stmt_to_callee): If call in noreturn; remove LHS.

From-SVN: r208901

re PR target/60675 ([aarch64] internal compiler error: Max. number of generated reload insns per insn is achieved (90))

2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>

PR target/60675
* lra-assigns.c (find_hard_regno_for): Remove unavailable hard
regs from checking multi-reg pseudos.

2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>

PR target/60675
* gcc.target/aarch64/pr60675.C: New.

From-SVN: r208900

simple-object.c (simple_object_internal_read): Handle EINTR and short reads.

2014-03-28  Richard Biener  <rguenther@suse.de>

libiberty/
* simple-object.c (simple_object_internal_read): Handle
EINTR and short reads.

lto-plugin/
* lto-plugin.c (process_symtab): Handle EINTR and short reads.

From-SVN: r208898

ssa-dom-thread-4.c: Remove s390 special option.

2014-03-28  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* gcc.dg/tree-ssa/ssa-dom-thread-4.c: Remove s390 special
          option.
* lib/target-supports.exp: Return true for s390
          in check_effective_logical_op_short_circuit.

From-SVN: r208897

Fix A12 multilib rule.

This affects only arm-none-eabi targets and those using t-aprofile in
their multilib lists. The problem here is that when the A12 support
was added, we mistakenly added this to the MULTILIB_MATCHES rule for
the A15 rather than putting out a separate line for this.

Fixed thusly and verified that the correct multilibs are now chosen.

     * config/arm/t-aprofile (MULTILIB_MATCHES): Correct A12 rule.

From-SVN: r208896

rs6000.c (fusion_gpr_load_p): Refuse optimization if it would clobber the stack pointer, even temporarily.

* config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
if it would clobber the stack pointer, even temporarily.

From-SVN: r208895

avx512f-vshuff32x4-2.c: Fix initialization of second source operand.

gcc/testsuite
* gcc.target/i386/avx512f-vshuff32x4-2.c: Fix initialization
  of second source operand.
* gcc.target/i386/avx512f-vshuff64x2-2.c: Ditto.
* gcc.target/i386/avx512f-vshufi32x4-2.c: Ditto.
* gcc.target/i386/avx512f-vshufi64x2-2.c: Ditto.

From-SVN: r208894

re PR ipa/60315 (template constructor switch optimization)

PR ipa/60315
* g++.dg/torture/pr60315.C: Add -std=c++11 to dg-options.

From-SVN: r208893

* mode-switching.c: Make small adjustments to the top comment.

From-SVN: r208892

cilk-plus-dg.exp: New.

2014-03-28  Tobias Burnus  <burnus@net-b.de>

        * lib/cilk-plus-dg.exp: New.
        * g++.dg/cilk-plus/cilk-plus.exp: Use it.
        * gcc.dg/cilk-plus/cilk-plus.exp: Use it.

From-SVN: r208889

re PR other/59545 (Signed integer overflow issues)

PR other/59545
compiler: Fix bogus invalid static_cast to Unary_expression.

From-SVN: r208888

Daily bump.

From-SVN: r208886

re PR fortran/60522 (WHERE construct causes an ICE in gfc_trans_where_2)

2014-04-27  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/60522
* frontend-passes.c (cfe_code):  Do not walk subtrees
for WHERE.

2014-04-27  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/60522
* gfortran.dg/where_4.f90:  New test case.

From-SVN: r208883

extend.texi (PowerPC AltiVec/VSX Built-in Functions): Document use of vec_xxsldwi and vec_xxpermdi builtins.

2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>

* doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
Document use of vec_xxsldwi and vec_xxpermdi builtins.

From-SVN: r208880

re PR fortran/58880 ([OOP] ICE on valid with FINAL function and type extension)

2014-03-27  Tobias Burnus  <burnus@net-b.de>

        PR fortran/58880
        * trans-expr.c (gfc_conv_scalar_to_descriptor): Fix handling
        of nonpointers.

2014-03-27  Tobias Burnus  <burnus@net-b.de>

        PR fortran/58880
        * gfortran.dg/finalize_24.f90: New.

From-SVN: r208879

re PR testsuite/60672 (FAIL: g++.dg/cpp1y/auto-fn25.C -std=gnu++1y  (test for errors, line 7))

[gcc]
2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>

* config/rs6000/constraints.md (wD constraint): New constraint to
match the constant integer to get the top DImode/DFmode out of a
vector in a VSX register.

* config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
match the constant integer to get the top DImode/DFmode out of a
vector in a VSX register.

* config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
for ISA 2.07.

* config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
vbpermq builtins.

* config/rs6000/rs6000.c (rs6000_debug_reg_global): If
-mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.

* config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
Optimize vec_extract of 64-bit values, where the value being
extracted is in the top word, where we can use scalar
instructions.  Add direct move and store support.  Combine the big
endian/little endian vector select load support into a single
insn.
(vsx_extract_<mode>_internal1): Likewise.
(vsx_extract_<mode>_internal2): Likewise.
(vsx_extract_<mode>_load): Likewise.
(vsx_extract_<mode>_store): Likewise.
(vsx_extract_<mode>_zero): Delete, big and little endian insns are
combined into vsx_extract_<mode>_load.
(vsx_extract_<mode>_one_le): Likewise.

* config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
define the top 64-bit vector element.

* doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
constraint.

PR target/60672
* config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
enable use of xxsldwi and xxpermdi builtin functions.
(vec_xxpermdi): Likewise.

* doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
Document use of vec_xxsldwi and vec_xxpermdi builtins.

[gcc/testsuite]
2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>

* gcc.target/powerpc/p8vector-vbpermq.c: New test to test the
vbpermq builtin.

* gcc.target/powerpc/vsx-extract-1.c: New test to test VSX
vec_select optimizations.
* gcc.target/powerpc/vsx-extract-2.c: Likewise.
* gcc.target/powerpc/vsx-extract-3.c: Likewise.

PR target/60672
* gcc.target/powerpc/pr60676.c: New file, make sure xxsldwi and
xxpermdi builtins are supported.

From-SVN: r208877

re PR rtl-optimization/60650 ([ARM] LRA ICE in assign_by_spills)

2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/60650
* lra-asign.c (find_hard_regno_for, spill_for): Add parameter
first_p.  Use it.
(find_spills_for): New.
(assign_by_spills): Pass the new parameter to find_hard_regno_for.
Spill all pseudos on the second iteration.

2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/60650
* gcc.target/arm/pr60650.c: New.

From-SVN: r208876

* doc/html/*: Regenerate.

From-SVN: r208875

containers.xml (associative.bitset.size_variable): Fix bad s/part/chapter/ substitutions.

* doc/xml/manual/containers.xml (associative.bitset.size_variable):
Fix bad s/part/chapter/ substitutions.
* doc/xml/manual/io.xml (std.io): Likewise.
* doc/xml/manual/numerics.xml (std.numerics.generalized_ops): Likewise.
* doc/xml/manual/strings.xml (strings.string.Cstring): Likewise.

From-SVN: r208874

faq.xml (faq): Refer to clauses instead of chapters.

* doc/xml/faq.xml (faq): Refer to clauses instead of chapters.
* doc/xml/manual/appendix_contributing.xml (contrib.design_notes):
Likewise.
* doc/xml/manual/backwards_compatibility.xml (backwards.third):
Likewise.
* doc/xml/manual/test.xml (test.organization.layout): Likewise.

From-SVN: r208873

* doc/xml/manual/io.xml (std.io.objects): Additional markup.

From-SVN: r208872

re PR libstdc++/60612 (Throwing exception, catching and rethrowing (std::exception_ptr) in destructor leads to segfault)

PR libstdc++/60612
* libsupc++/eh_ptr.cc: Assert __cxa_dependent_exception layout is
compatible with __cxa_exception.
* libsupc++/unwind-cxx.h (__cxa_dependent_exception): Add padding.
Fix typos in comments.
* testsuite/18_support/exception_ptr/60612-terminate.cc: New.
* testsuite/18_support/exception_ptr/60612-unexpected.cc: New.

From-SVN: r208871

re PR c/50347 (unexpected -Wconversion error from gcc builtin)

PR c/50347
* doc/extend.texi (ffs Builtins): Change unsigned types to signed
types.

From-SVN: r208870

config.host: Append t-floattodi to tmake_file depending on host_address.

2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* config.host: Append t-floattodi to tmake_file depending on
host_address.

From-SVN: r208868

s390.c (s390_can_use_return_insn): Check for call-saved FPRs on 31 bit.

2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* config/s390/s390.c (s390_can_use_return_insn): Check for
call-saved FPRs on 31 bit.

2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* gcc.target/s390/20140327-1.c: New testcase.

From-SVN: r208867

Makefile.def (dependencies): Make all-ld depend on all-binutils for WINDRES_FOR_TARGET in default-manifest.o...

* Makefile.def (dependencies): Make all-ld depend on all-binutils
for WINDRES_FOR_TARGET in default-manifest.o rule.
* Makefile.in: Regenerate.

From-SVN: r208865

re PR middle-end/60682 ([OpenMP] ICE on an assignment of local variable inside SIMD loop)

PR middle-end/60682
* omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
if they need regimplification, just drop them instead of
calling gimple_regimplify_operands on them.

* g++.dg/gomp/pr60682.C: New test.

From-SVN: r208864

pr60092.c: Remove default dg-skip-if arguments.

* gcc.dg/torture/pr60092.c: Remove default dg-skip-if arguments.

From-SVN: r208863

[PR 60580, AArch64] Fix __attribute__ ((optimize("no-omit-frame-pointer")))

The implementation of -m[no-]omit-leaf-frame-pointer and
-f[no-]omit-frame-pointer in the AArch64 target does not behave
correctly in the presence of __attribute__ optimize.

This patch adjusts the implementation to work in a similar fashion to
the same functionality in the i386 target.

The problem occurs because the current implementation uses a global
'faked_omit_frame_pointer' to retain the original value of
flag_omit_frame_pointer.  The global does not form part of the
optimization save state.

This solution removes the global and instead tracks required behaviour
using only flag_omit_frame_pointer and flag_omit_leaf_frame_pointer.
These two form part of the optimziation save state and target save
state respectively.

The additional complication for AArch64 is that the PCS requires that
given -fno-omit-frame-pointer -momit-leave-frame-pointer, a leaf
function that kills LR must create a frame record.  This is readily
handled in aarch64_frame_pointer_required().  I've dropped logic in
aarch64_can_eliminate() that attempts to detect this scenario since it

From-SVN: r208862

* ipa-inline.c (early_inliner): Update node's inline info.

From-SVN: r208861

dojump.c (do_compare_rtx_and_jump): Sets correct probability for compiler inserted conditional jumps for NAN float...

* dojump.c (do_compare_rtx_and_jump): Sets correct probability for
compiler inserted conditional jumps for NAN float check.
* gcc.dg/predict-8.c: New test.

From-SVN: r208860

Daily bump.

From-SVN: r208859

futex_bits.h: Include errno.h.

* config/linux/futex_bits.h: Include errno.h.
(sys_futex0): If syscall returns -1, return -errno rather than
-1.

From-SVN: r208855

re PR c++/52369 (Const-qualified non-class base member and defaulted default constructor)

2014-03-26  Fabien Chene  <fabien@gcc.gnu.org>
        PR c++/52369
        * cp/method.c (walk_field_subobs): improve the diagnostic
locations for both REFERENCE_TYPEs and non-static const members.
* cp/init.c (diagnose_uninitialized_cst_or_ref_member): use %q#D
instead of %qD to be consistent with the c++11 diagnostic.

2014-03-26  Fabien Chene  <fabien@gcc.gnu.org>

        PR c++/52369
        * g++.dg/init/const10.C: New.
* g++.dg/init/const11.C: New.
* g++.dg/init/pr25811.C: Adjust.
* g++.dg/init/pr29043.C: Likewise.
* g++.dg/init/pr43719.C: Likewise.
* g++.dg/init/pr44086.C: Likewise.
* g++.dg/init/ctor8.C: Likewise.
* g++.dg/init/uninitialized1.C: Likewise.

From-SVN: r208854

* libitm.texi (Index): Rename to Library Index.

From-SVN: r208853

re PR fortran/34928 (Extension: volatile common blocks)

        PR fortran/34928
        * fortran.texi: Document Volatile COMMON as not supported.

From-SVN: r208852

Add libcilkrts/libcilkrts.spec.in, missing from r208847

From-SVN: r208851

ubsan.h (ubsan_create_data): Change second argument's type to const location_t *.

* ubsan.h (ubsan_create_data): Change second argument's type
to const location_t *.
* ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
_("<unknown>").
(ubsan_create_data): Change second argument to const location_t *PLOC.
Create Loc field whenever PLOC is non-NULL.
(ubsan_instrument_unreachable, ubsan_expand_null_ifn,
ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
callers.
c-family/
* c-ubsan.c (ubsan_instrument_division, ubsan_instrument_shift,
ubsan_instrument_vla, ubsan_instrument_return): Adjust
ubsan_create_data callers.

From-SVN: r208849

re PR other/59545 (Signed integer overflow issues)

PR other/59545
* real.c (real_to_integer2): Change type of low to UHWI.

From-SVN: r208848

gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.

gcc/
2014-03-26  Tobias Burnus  <burnus@net-b.de>

        * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
        (CILK_SELF_SPECS): New define.
        (driver_self_specs): Use it.

libcilkrts/
2014-03-26  Tobias Burnus  <burnus@net-b.de>

        * libcilkrts.spec.in: New.
        * Makefile.am: Handle libcilkrts.spec.
        * configure.ac: Determine link options for libcilkrts.spec.
        * Makefile.in: Regenerate.
        * configure: Regenerate.
        * aclocal.m4: Regenerate.

From-SVN: r208847

MAINTAINERS (Write After Approval): Add myself.

        * MAINTAINERS (Write After Approval): Add myself.

From-SVN: r208846

re PR c++/60566 (r208573 omits needed thunks)

PR c++/60566
PR c++/58678
* class.c (build_vtbl_initializer): Handle abstract dtors here.
* search.c (get_pure_virtuals): Not here.

From-SVN: r208845

tree-pretty-print.c (percent_K_format): Implement special case for LTO and its stripped down BLOCK tree.

2014-03-26  Richard Biener  <rguenther@suse.de>

* tree-pretty-print.c (percent_K_format): Implement special
case for LTO and its stripped down BLOCK tree.

From-SVN: r208843

re PR sanitizer/56781 (boostrap-asan failure: fixincl fails to link (missing -lasan))

PR sanitizer/56781
* Makefile.def: Set bootstrap=true; for host fixincludes.
* configure.ac: Don't bootstrap host fixincludes unless
--with-build-config=bootstrap-{a,ub}san.
* Makefile.in: Regenerated.
* configure: Regenerated.

From-SVN: r208842

re PR sanitizer/60636 (ubsan doesn't instrument signed integer ABS_EXPR)

PR sanitizer/60636
* ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.

* c-c++-common/ubsan/pr60636.c: New test.

From-SVN: r208841

tree-vrp.c (simplify_internal_call_using_ranges): If only one range is range_int_cst_p...

* tree-vrp.c (simplify_internal_call_using_ranges): If only
one range is range_int_cst_p, but not both, at least optimize
addition/subtraction of 0 and multiplication by 0 or 1.
* gimple-fold.c (gimple_fold_call): Fold
IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
(gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
INTEGER_CSTs, try to fold at least x * 0 and y - y.

From-SVN: r208840

* g++.dg/torture/pr60315.C: Remove duplication.

From-SVN: r208839

re PR rtl-optimization/60452 (wrong code at -O1 with large offsets in the frame)

PR rtl-optimization/60452
* rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
<case REG>: Return 1 for invalid offsets from the frame pointer.

From-SVN: r208837

re PR c/37428 (GNU VLA-in-structure extension is undocumented)

PR c/37428
* doc/extend.texi (C Extensions): Mention variable-length arrays in
a structure/union.

From-SVN: r208836

re PR c/39525 ([easy to fix bug] Docs for C99 "designated initializers" isn't clear what happens to omitted field members)

PR c/39525
* doc/extend.texi (Designated Inits): Describe what happens to omitted
field members.

From-SVN: r208835

re PR other/59545 (Signed integer overflow issues)

PR other/59545
* ira-color.c (update_conflict_hard_regno_costs): Perform the
multiplication in unsigned type.

From-SVN: r208834

* doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.

From-SVN: r208833

* doc/contrib.texi: Add myself as Andes nds32 port contributor.

From-SVN: r208832

re PR ipa/60315 (template constructor switch optimization)

PR ipa/60315
* cif-code.def (UNREACHABLE) New code.
* ipa-inline.c (inline_small_functions): Skip edges to __builtlin_unreachable.
(estimate_edge_growth): Allow edges to __builtlin_unreachable.
* ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
predicate to __bulitin_unreachable.
(set_cond_stmt_execution_predicate): Fix issue when invert_tree_comparison
returns ERROR_MARK.
* ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
propagate to inline clones.
* cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
to unreachable.
* ipa-cp.c (create_specialized_node): Be ready for new node to appear.
* cgraphclones.c (cgraph_clone_node): If call destination is already
ureachable, do not redirect it back.
* tree-inline.c (fold_marked_statements): Hanlde calls becoming
unreachable.

From-SVN: r208831

ipa-pure-const.c (propagate_pure_const, [...]): Do not modify inline clones.

* ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
Do not modify inline clones.

From-SVN: r208830

Daily bump.

From-SVN: r208829

i386.md (general_sext_operand): New mode attr.

* config/i386/i386.md (general_sext_operand): New mode attr.
(addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
don't generate (sign_extend (const_int)).
(*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
operands[2].  Use We constraint instead of <i> and <general_sext_operand>
predicate instead of <general_operand>.
(*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
* config/i386/constraints.md (We): New constraint.
* config/i386/predicates.md (x86_64_sext_operand,
sext_operand): New predicates.

From-SVN: r208824

re PR libstdc++/60658 (std::atomic<T*> is unexpectedly not lock-free)

PR libstdc++/60658
* include/bits/atomic_base.h (__atomic_base<_PTp*>::is_lock_free()):
Use sizeof pointer type not the element type.
* testsuite/29_atomics/atomic/60658.cc: New.

From-SVN: r208819

re PR ipa/60600 (ICE in ipa_get_indirect_edge_target_1)

2014-03-25  Martin Jambor  <mjambor@suse.cz>

PR ipa/60600
* ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
inconsistent devirtualizations to __builtin_unreachable.

testsuite/
* g++.dg/ipa/pr60600.C: New test.

From-SVN: r208818

re PR c++/60375 ([c++11] ICE with invalid use of lambda)

PR c++/60375
* parser.c (cp_parser_lambda_expression): Don't parse the body of
a lambda in unevaluated context.

From-SVN: r208817

re PR c++/60628 ([c++11] ICE initializing array of auto)

PR c++/60628
* decl.c (create_array_type_for_decl): Complain about array of auto.

From-SVN: r208816

re PR c/35449 (extended asm documentation wrong)

PR c/35449
* doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.

From-SVN: r208815

[PATCH AArch64] Fix aarch64_simd_valid_immediate for Bigendian

gcc/

* config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
order of elements for big-endian.

From-SVN: r208814

re PR testsuite/58013 (main() not generated in assembler output)

PR testsuite/58013
* g++.dg/opt/pr56999.C: Pop hidden visibility.

From-SVN: r208813

re PR lto/60635 (ICE when mixing C and Fortran lto1: error: use operand missing for stmt)

2014-03-25  Richard Biener  <rguenther@suse.de>

PR middle-end/60635
* gimplify-me.c (gimple_regimplify_operands): Update the
re-gimplifed stmt.

* gfortran.dg/lto/pr60635_0.f90: New testcase.
* gfortran.dg/lto/pr60635_1.c: Likewise.

From-SVN: r208811

re PR ipa/59176 (ICE edge points to wrong declaration / verify_cgraph_node failed)

2014-03-25  Martin Jambor  <mjambor@suse.cz>

PR ipa/59176
* lto-cgraph.c (lto_output_node): Stream body_removed flag.
(lto_output_varpool_node): Likewise.
(input_overwrite_node): Likewise.
(input_varpool_node): Likewise.

From-SVN: r208809

lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like OPT_fpie.

2014-03-25  Richard Biener  <rguenther@suse.de>

* lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like
OPT_fpie.
(run_gcc): Likewise.

From-SVN: r208807

re PR c++/60331 (ICE with OpenMP #pragma omp declare reduction in template class)

PR c++/60331
* semantics.c (potential_constant_expression_1): Handle
DECL_EXPR.

* testsuite/libgomp.c++/udr-11.C: New test.
* testsuite/libgomp.c++/udr-12.C: New test.
* testsuite/libgomp.c++/udr-13.C: New test.
* testsuite/libgomp.c++/udr-14.C: New test.
* testsuite/libgomp.c++/udr-15.C: New test.
* testsuite/libgomp.c++/udr-16.C: New test.
* testsuite/libgomp.c++/udr-17.C: New test.
* testsuite/libgomp.c++/udr-18.C: New test.
* testsuite/libgomp.c++/udr-19.C: New test.

From-SVN: r208806

combine.c (simplify_compare_const): Add MODE argument.

* combine.c (simplify_compare_const): Add MODE argument.
Handle mode_width 0 as very large mode_width.
(try_combine, simplify_comparison): Adjust callers.

From-SVN: r208805

cselib.c (cselib_hash_rtx): Perform addition in unsigned type to avoid signed integer overflow.

* cselib.c (cselib_hash_rtx): Perform addition in unsigned
type to avoid signed integer overflow.
* explow.c (plus_constant): Likewise.

From-SVN: r208804

generic.texi: Correct typos.

2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>

* doc/generic.texi: Correct typos.

From-SVN: r208803

Daily bump.

From-SVN: r208802

re PR c++/60627 ([c++1y] ICE in explicit template instantiation containing auto parameter)

Fix PR c++/60627

PR c++/60627
* parser.c (cp_parser_parameter_declaration_clause): Prevent 'auto' from
introducing an implicit function template parameter within an explicit
instantiation.

PR c++/60627
* g++.dg/cpp1y/pr60627.C: New testcase.

From-SVN: r208799

invoke.texi (-flto): Expand section about using static libraries with LTO.

2014-03-24  Tobias Burnus  <burnus@net-b.de>

        * doc/invoke.texi (-flto): Expand section about
        using static libraries with LTO.

From-SVN: r208798

baseline_symbols.txt: Update.

* config/abi/post/powerpc64-linux-gnu/32/baseline_symbols.txt: Update.
* config/abi/post/powerpc64-linux-gnu/baseline_symbols.txt: Likewise.
* config/abi/post/s390-linux-gnu/baseline_symbols.txt: Likewise.
* config/abi/post/s390x-linux-gnu/baseline_symbols.txt: Likewise.
* config/abi/post/i486-linux-gnu/baseline_symbols.txt: Likewise.
* config/abi/post/x86_64-linux-gnu/32/baseline_symbols.txt: Likewise.
* config/abi/post/x86_64-linux-gnu/baseline_symbols.txt: Likewise.
* config/abi/post/powerpc-linux-gnu/baseline_symbols.txt: Likewise.
* config/abi/post/i386-linux-gnu/baseline_symbols.txt: Likewise.
* config/abi/post/hppa-linux-gnu/baseline_symbols.txt: Remove TLS
entries.

From-SVN: r208797

re PR rtl-optimization/60501 (LRA emits add patterns which might clobber cc)

2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

PR rtl-optimization/60501
* optabs.def (addptr3_optab): New optab.
* optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
* doc/md.texi ("addptrm3"): Document new RTL standard expander.
* expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.

* lra.c (emit_add3_insn): Use the addptr pattern if available.

* config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.

From-SVN: r208796

avx512fintrin.h: Define _mm512_set1_ps and _mm512_set1_pd.

2014-03-24  Ulrich Drepper  <drepper@gmail.com>

        * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
        _mm512_set1_pd.

From-SVN: r208794

avxintrin.h (_mm256_undefined_si256): Define.

2014-03-24  Ulrich Drepper  <drepper@gmail.com>

        * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
        (_mm256_undefined_ps): Define.
        (_mm256_undefined_pd): Define.
        * config/i386/emmintrin.h (_mm_undefined_si128): Define.
        (_mm_undefined_pd): Define.
        * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
        * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
        (_mm512_undefined_ps): Define.
        (_mm512_undefined_pd): Define.
        Use _mm*_undefined_*.
        * config/i386/avx2intrin.h: Use _mm*_undefined_*.

From-SVN: r208793

re PR c++/60574 ([c++1y] ICE with deduced return type in virtual function and LTO)

PR c++/60574
* decl.c (grokdeclarator): Change permerror about 'virtual auto'
to error.

From-SVN: r208792

* sv.po: Update.

From-SVN: r208791

[AArch64] Logical vector shift right conformance

gcc/

* config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
(lshr_simd): DI mode added.
* config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
(aarch64_ushr_simddi): Likewise.
* config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
* config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
(vshrd_n_u64): Likewise.

gcc/testsuite/

* gcc.target/aarch64/ushr64_1.c: New testcase.

From-SVN: r208789

[AArch64 testsuite] Make vect-abs.c C99 compliant

* gcc.target/aarch64/vect-abs.c (dg-options): Add -std=c99.
(LONG_LONG): Use LLONG.
(set_rvector_long): Explicitly return void.
(set_vector_long): Likewise.
(check_vector_long): Likewise.

From-SVN: r208788

div-by-zero-4.c: Don't include limits.h.

* c-c++-common/ubsan/div-by-zero-4.c: Don't include limits.h.  Define
INT_MIN.
* c-c++-common/ubsan/overflow-1.c: Check for unwanted output.
* c-c++-common/ubsan/overflow-add-1.c: Likewise.
* c-c++-common/ubsan/overflow-mul-1.c: Likewise.
* c-c++-common/ubsan/overflow-mul-3.c: Likewise.
* c-c++-common/ubsan/overflow-negate-2.c: Likewise.
* c-c++-common/ubsan/overflow-sub-1.c: Likewise.
* c-c++-common/ubsan/pr59503.c: Likewise.
* c-c++-common/ubsan/pr60613-1.c: Likewise.
* c-c++-common/ubsan/save-expr-1.c: Likewise.
* c-c++-common/ubsan/shift-3.c: Likewise.
* c-c++-common/ubsan/shift-6.c: Likewise.
* c-c++-common/ubsan/undefined-1.c: Likewise.
* c-c++-common/ubsan/vla-2.c: Likewise.
* c-c++-common/ubsan/vla-3.c: Likewise.
* c-c++-common/ubsan/vla-4.c: Likewise.
* g++.dg/ubsan/cxx11-shift-1.C: Likewise.
* g++.dg/ubsan/return-2.C: Likewise.

From-SVN: r208787

Have s-macro_list depend on cc1

* Makefile.in (s-macro_list): Depend on cc1.

From-SVN: r208786

pr59003.c: Add -fno-common to dg-options on hppa*-*-hpux*.

* gcc.dg/tree-prof/pr59003.c: Add -fno-common to dg-options on
hppa*-*-hpux*.

From-SVN: r208783

pr60092.c: Skip on 32-bit hpux.

* gcc.dg/torture/pr60092.c: Skip on 32-bit hpux.

From-SVN: r208782