AVX-512. Extend gather insn patterns.

gcc/
* config/i386/sse.md
(define_expand "<avx512>_gathersi<mode>"): Rename from
"avx512f_gathersi<mode>".
(define_insn "*avx512f_gathersi<mode>"): Use VI48F.
(define_insn "*avx512f_gathersi<mode>_2"): Ditto.
(define_expand "<avx512>_gatherdi<mode>"): Rename from
"avx512f_gatherdi<mode>".
(define_insn "*avx512f_gatherdi<mode>"): Use VI48F.
(define_insn "*avx512f_gatherdi<mode>_2"): Use VI48F, add 128/256-bit
wide versions.
(define_expand "<avx512>_scattersi<mode>"): Rename from
"avx512f_scattersi<mode>".
(define_insn "*avx512f_scattersi<mode>"): Use VI48F.
(define_expand "<avx512>_scatterdi<mode>"): Rename from
"avx512f_scatterdi<mode>".
(define_insn "*avx512f_scatterdi<mode>"): Use VI48F.

Co-Authored-By: Andrey Turetskiy <andrey.turetskiy@intel.com>
Co-Authored-By: Anna Tikhonova <anna.tikhonova@intel.com>
Co-Authored-By: Ilya Tocar <ilya.tocar@intel.com>
Co-Authored-By: Ilya Verbin <ilya.verbin@intel.com>
Co-Authored-By: Kirill Yukhin <kirill.yukhin@intel.com>
Co-Authored-By: Maxim Kuznetsov <maxim.kuznetsov@intel.com>
Co-Authored-By: Michael Zolotukhin <michael.v.zolotukhin@intel.com>
From-SVN: r215201

ira.h (ira_finish_once): Delete.

gcc/
* ira.h (ira_finish_once): Delete.
* ira-int.h (target_ira_int::~target_ira_int): Declare.
(target_ira_int::free_ira_costs): Likewise.
(target_ira_int::free_register_move_costs): Likewise.
(ira_finish_costs_once): Delete.
* ira.c (free_register_move_costs): Replace with...
(target_ira_int::free_register_move_costs): ...this new function.
(target_ira_int::~target_ira_int): Define.
(ira_init): Call free_register_move_costs as a member function rather
than a global function.
(ira_finish_once): Delete.
* ira-costs.c (free_ira_costs): Replace with...
(target_ira_int::free_ira_costs): ...this new function.
(ira_init_costs): Call free_ira_costs as a member function rather
than a global function.
(ira_finish_costs_once): Delete.
* target-globals.c (target_globals::~target_globals): Call the
target_ira_int destructor.
* toplev.c: Include lra.h.
(finalize): Call lra_finish_once rather than ira_finish_once.

From-SVN: r215200

Daily bump.

From-SVN: r215199

common.opt (flto-odr-type-merging): New flag.

* common.opt (flto-odr-type-merging): New flag.
* ipa-deivrt.c (hash_type_name): Use ODR names for hasing if availale.
(types_same_for_odr): Likewise.
(odr_subtypes_equivalent_p): Likewise.
(add_type_duplicate): Do not walk type variants.
(register_odr_type): New function.
* ipa-utils.h (register_odr_type): Declare.
(odr_type_p): New function.
* langhooks.c (lhd_set_decl_assembler_name): Do not compute
TYPE_DECLs
* doc/invoke.texi (-flto-odr-type-merging): Document.
* tree.c (need_assembler_name_p): Compute ODR names when asked
for it.
* tree.h (DECL_ASSEMBLER_NAME): Update comment.

* lto.c (lto_read_decls): Register ODR types.

From-SVN: r215196

Also turn off OPTION_MASK_ABI_X32 for -m16

PR target/63228
* config/i386/i386.c (ix86_option_override_internal): Also turn
off OPTION_MASK_ABI_X32 for -m16.

From-SVN: r215194

rs6000.md (rs6000_mftb_<mode>): Use mode iterator GPR instead of P.

2014-09-11  Segher Boessenkool  <segher@kernel.crashing.org>

* config/rs6000/rs6000.md (rs6000_mftb_<mode>): Use mode iterator
GPR instead of P.

From-SVN: r215193

re PR target/58757 (Advertise the lack of denormal support on alpha without -mieee)

2014-09-11  Marc Glisse  <marc.glisse@inria.fr>

PR target/58757
gcc/c-family/
* c-cppbuiltin.c (builtin_define_float_constants): Correct
__*_DENORM_MIN__ without denormals.
gcc/
* ginclude/float.h (FLT_TRUE_MIN, DBL_TRUE_MIN, LDBL_TRUE_MIN):
Directly forward to __*_DENORM_MIN__.
gcc/testsuite/
* gcc.dg/c11-true_min-1.c: New testcase.

From-SVN: r215191

Introduce LABEL_REF_LABEL

gcc/ChangeLog:
2014-09-11  David Malcolm  <dmalcolm@redhat.com>

* rtl.h (LABEL_REF_LABEL): New macro.

* alias.c (rtx_equal_for_memref_p): Use LABEL_REF_LABEL in place
of XEXP (, 0), where we know that we have a LABEL_REF.
* cfgbuild.c (make_edges): Likewise.
(purge_dead_tablejump_edges): Likewise.
* cfgexpand.c (convert_debug_memory_address): Likewise.
* cfgrtl.c (patch_jump_insn): Likewise.
* combine.c (distribute_notes): Likewise.
* cse.c (hash_rtx_cb): Likewise.
(exp_equiv_p): Likewise.
(fold_rtx): Likewise.
(check_for_label_ref): Likewise.
* cselib.c (rtx_equal_for_cselib_1): Likewise.
(cselib_hash_rtx): Likewise.
* emit-rtl.c (mark_label_nuses): Likewise.
* explow.c (convert_memory_address_addr_space): Likewise.
* final.c (output_asm_label): Likewise.
(output_addr_const): Likewise.
* gcse.c (add_label_notes): Likewise.
* genconfig.c (walk_insn_part): Likewise.
* genrecog.c (validate_pattern): Likewise.
* ifcvt.c (cond_exec_get_condition): Likewise.
(noce_emit_store_flag): Likewise.
(noce_get_alt_condition): Likewise.
(noce_get_condition): Likewise.
* jump.c (maybe_propagate_label_ref): Likewise.
(mark_jump_label_1): Likewise.
(redirect_exp_1): Likewise.
(rtx_renumbered_equal_p): Likewise.
* lra-constraints.c (operands_match_p): Likewise.
* reload.c (operands_match_p): Likewise.
(find_reloads): Likewise.
* reload1.c (set_label_offsets): Likewise.
* reorg.c (get_branch_condition): Likewise.
* rtl.c (rtx_equal_p_cb): Likewise.
(rtx_equal_p): Likewise.
* rtlanal.c (reg_mentioned_p): Likewise.
(rtx_referenced_p): Likewise.
(get_condition): Likewise.
* sched-vis.c (print_value): Likewise.
* varasm.c (const_hash_1): Likewise.
(compare_constant): Likewise.
(const_rtx_hash_1): Likewise.
(output_constant_pool_1): Likewise.

From-SVN: r215190

htm.md (tabort, [...]): Use xor instead of minus.

2014-09-11  Segher Boessenkool  <segher@kernel.crashing.org>

* config/rs6000/htm.md (tabort, tabortdc, tabortdci, tabortwc,
tabortwci, tbegin, tcheck, tend, trechkpt, treclaim, tsr): Use xor
instead of minus.
* config/rs6000/vector.md (cr6_test_for_zero_reverse,
cr6_test_for_lt_reverse): Ditto.

From-SVN: r215187

re PR c++/61489 (Wrong warning with -Wmissing-field-initializers.)

2014-09-11  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/61489
* doc/invoke.texi ([-Wmissing-field-initializers]): Update.

/cp
2014-09-11  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/61489
* typeck2.c (process_init_constructor_record): Do not warn about
missing field initializer if EMPTY_CONSTRUCTOR_P (init).

/testsuite
2014-09-11  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/61489
* g++.dg/warn/Wmissing-field-initializers-1.C: New.
* g++.old-deja/g++.other/warn5.C: Adjust.

From-SVN: r215186

[AArch64] Simplify vreinterpret for float64x1_t using casts.

* config/aarch64/aarch64-builtins.c (aarch64_types_unop_su_qualifiers,
TYPES_REINTERP_SU, aarch64_types_unop_sp_qualifiers, TYPE_REINTERP_SP,
aarch64_types_unop_us_qualifiers, TYPES_REINTERP_US,
aarch64_types_unop_ps_qualifiers, TYPES_REINTERP_PS, BUILTIN_VD):
Delete.

(aarch64_fold_builtin): Remove all reinterpret cases.

* config/aarch64/aarch64-protos.h (aarch64_simd_reinterpret): Delete.

* config/aarch64/aarch64-simd-builtins.def (reinterpret*) : Delete.

* config/aarch64/aarch64-simd.md (aarch64_reinterpretv8qi<mode>,
aarch64_reinterpretv4hi<mode>, aarch64_reinterpretv2si<mode>,
aarch64_reinterpretv2sf<mode>, aarch64_reinterpretdi<mode>,
aarch64_reinterpretv1df<mode>, aarch64_reinterpretv16qi<mode>,
aarch64_reinterpretv8hi<mode>, aarch64_reinterpretv4si<mode>,
aarch64_reinterpretv4sf<mode>, aarch64_reinterpretv2di<mode>,
aarch64_reinterpretv2df<mode>): Delete.

* config/aarch64/aarch64.c (aarch64_simd_reinterpret): Delete.

* config/aarch64/arm_neon.h (vreinterpret_p8_f64,
vreinterpret_p16_f64, vreinterpret_f32_f64, vreinterpret_f64_f32,
vreinterpret_f64_p8, vreinterpret_f64_p16, vreinterpret_f64_s8,
vreinterpret_f64_s16, vreinterpret_f64_s32, vreinterpret_f64_u8,
vreinterpret_f64_u16, vreinterpret_f64_u32, vreinterpret_s64_f64,
vreinterpret_u64_f64, vreinterpret_s8_f64, vreinterpret_s16_f64,
vreinterpret_s32_f64, vreinterpret_u8_f64, vreinterpret_u16_f64,
vreinterpret_u32_f64): Use cast.

* config/aarch64/iterators.md (VD_RE): Delete.

From-SVN: r215180

[AArch64] Replace temporary inline assembler for vset_lane

* config/aarch64/arm_neon.h (aarch64_vset_lane_any): New (*2).
(vset_lane_f32, vset_lane_f64, vset_lane_p8, vset_lane_p16,
vset_lane_s8, vset_lane_s16, vset_lane_s32, vset_lane_s64,
vset_lane_u8, vset_lane_u16, vset_lane_u32, vset_lane_u64,
vsetq_lane_f32, vsetq_lane_f64, vsetq_lane_p8, vsetq_lane_p16,
vsetq_lane_s8, vsetq_lane_s16, vsetq_lane_s32, vsetq_lane_s64,
vsetq_lane_u8, vsetq_lane_u16, vsetq_lane_u32, vsetq_lane_u64):
Replace inline assembler with __aarch64_vset_lane_any.

From-SVN: r215179

[AArch64 Testsuite] Add execution test of vset(q?)_lane intrinsics.

* gcc.target/aarch64/vset_lane_1.c: New test.

From-SVN: r215177

[AArch64] Cheap fix for argument types of vmull_high_lane_{us}{16,32}

gcc/

* config/aarch64/arm_neon.h (vmull_high_lane_s16): Fix argument
types.
(vmull_high_lane_s32): Likewise.
(vmull_high_lane_u16): Likewise.
(vmull_high_lane_u32): Likewise.

From-SVN: r215176

re PR c++/63139 (Class-scope typedef overwrites typedef of previously defined class)

PR c++/63139
* pt.c (tsubst_pack_expansion): Simplify substitution into T....
(tsubst): Don't throw away PACK_EXPANSION_EXTRA_ARGS.

From-SVN: r215171

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

PR c++/58678
* ipa-devirt.c (ipa_devirt): Don't check DECL_COMDAT.

From-SVN: r215168

re PR libstdc++/63219 (Superfluous template parameter in match_result::format overload)

PR libstdc++/63219
* include/bits/regex.h (match_results::format): Remove stray template
parameter.
* include/bits/regex_compiler.h (_RegexTranslator::_RegexTranslator):
Remove parameter name to avoid -Wunused-parameter warning.
* include/bits/regex_executor.h (_State_info::_State_info): Reorder
mem-initializers to avoid -Wreorder warning.
* include/bits/regex_executor.tcc (_Executor::_M_word_boundary):
Remove parameter name to avoid -Wunused-parameter warning.
* include/bits/regex_scanner.tcc (_Scanner::_M_advance): Add braces
to avoid -Wempty-body warning when not in debug mode.

From-SVN: r215160

Fix declarations in some tests.

* gcc.dg/compat/struct-by-value-13_main.c (struct_by_value_13_x):
Fix declaration.
* gcc.dg/compat/struct-by-value-16a_main.c (struct_by_value_16a_x):
Fix declaration.
* gcc.dg/compat/struct-by-value-17a_main.c (struct_by_value_17a_x):
Fix declaration.
* gcc.dg/compat/struct-by-value-18a_main.c (struct_by_value_18a_x):
Fix declaration.

From-SVN: r215158

linux64.S: Emit .note.GNU-stack even when POWERPC64 is not defined.

* src/powerpc/linux64.S: Emit .note.GNU-stack even when
POWERPC64 is not defined.
* src/powerpc/linux64_closure.S: Likewise.  Also test _CALL_ELF == 2.

From-SVN: r215155

re PR target/63223 ([avr] Make jumptables work with -Wl,--section-start,.text=)

gcc/
PR target/63223
* config/avr/avr.md (*tablejump.3byte-pc): New insn.
(*tablejump): Restrict to !AVR_HAVE_EIJMP_EICALL.  Add void clobber.
(casesi): Expand to *tablejump.3byte-pc if AVR_HAVE_EIJMP_EICALL.
libgcc/
PR target/63223
* config/avr/libgcc.S (__tablejump2__): Rewrite to use RAMPZ, ELPM
and R24 as needed.  Make work for all devices and .text locations.
(__do_global_ctors, __do_global_dtors): Use word addresses.
(__tablejump__, __tablejump_elpm__): Remove functions.
* t-avr (LIB1ASMFUNCS): Remove _tablejump, _tablejump_elpm.
Add _tablejump2.
(XICALL, XIJMP): New macros.

From-SVN: r215152

AVX-512. Add vperm[it]2 insns support.

gcc/
* config/i386/sse.md
(define_expand "<avx512>_vpermi2var<VI48F:mode>3_maskz"): Rename from
"avx512f_vpermi2var<mode>3_maskz" and update mode iterator.
(define_expand "<avx512>_vpermi2var<VI2_AVX512VL:mode>3_maskz"):
New.
(define_insn "<avx512>_vpermi2var<VI48F:mode>3<sd_maskz_name>"): Rename
from "avx512f_vpermi2var<mode>3<sd_maskz_name>" and update mode
iterator.
(define_insn "<avx512>_vpermi2var<VI2_AVX512VL:mode>3<sd_maskz_name>"):
New.
(define_insn "<avx512>_vpermi2var<VI48F:mode>3_mask"): Rename from
"avx512f_vpermi2var<mode>3_mask" and update mode iterator.
(define_insn "<avx512>_vpermi2var<VI2_AVX512VL:mode>3_mask"): New.
(define_expand "<avx512>_vpermt2var<VI48F:mode>3_maskz"): Rename from
"avx512f_vpermt2var<mode>3_maskz" and update mode iterator.
(define_expand "<avx512>_vpermt2var<VI2_AVX512VL:mode>3_maskz"): New.
(define_insn "<avx512>_vpermt2var<VI48F:mode>3<sd_maskz_name>"): Rename
from "avx512f_vpermt2var<mode>3<sd_maskz_name>" and update mode
iterator.
(define_insn "<avx512>_vpermt2var<VI2_AVX512VL:mode>3<sd_maskz_name>"):
New.
(define_insn "<avx512>_vpermt2var<VI48F:mode>3_mask"): Rename from
"avx512f_vpermt2var<mode>3_mask" and update mode iterator.
(define_insn "<avx512>_vpermt2var<VI2_AVX512VL:mode>3_mask"): New.

Co-Authored-By: Andrey Turetskiy <andrey.turetskiy@intel.com>
Co-Authored-By: Anna Tikhonova <anna.tikhonova@intel.com>
Co-Authored-By: Ilya Tocar <ilya.tocar@intel.com>
Co-Authored-By: Ilya Verbin <ilya.verbin@intel.com>
Co-Authored-By: Kirill Yukhin <kirill.yukhin@intel.com>
Co-Authored-By: Maxim Kuznetsov <maxim.kuznetsov@intel.com>
Co-Authored-By: Michael Zolotukhin <michael.v.zolotukhin@intel.com>
From-SVN: r215151

varpool.c (varpool_node::ctor_useable_for_folding_p): Do not try to access removed nodes.

* varpool.c (varpool_node::ctor_useable_for_folding_p): Do not try
to access removed nodes.

From-SVN: r215150

re PR middle-end/63186 (Undefined .L* symbols because of fnsplit)

PR tree-optimization/63186
* ipa-split.c (test_nonssa_use): Skip nonforced labels.
(mark_nonssa_use): Likewise.
(verify_non_ssa_vars): Verify all header blocks for label
definitions.

* gcc.dg/pr63186.c: New testcase.

From-SVN: r215149

AVX-512. Extend vpermvar insn patterns.

gcc/
* config/i386/sse.md
(define_mode_attr avx2_avx512): Rename from avx2_avx512bw.
(define_mode_iterator VI48F_256_512): Extend to AVX-512VL.
(define_insn "<avx2_avx512>_permvar<mode><mask_name>"): Rename from
"<avx2_avx512f>_permvar<mode><mask_name>".
(define_insn "<avx512>_permvar<mode><mask_name>"): New.
(define_insn "<avx2_avx512>_ashrv<VI48_AVX512F_AVX512VL:mode><mask_name>"):
Rename from "<avx2_avx512f>_ashrv<mode><mask_name>".
(define_insn "<avx2_avx512>_ashrv<VI2_AVX512VL:mode><mask_name>"):
Ditto.
(define_insn "<avx2_avx512>_<shift_insn>v<VI48_AVX512F:mode><mask_name>"):
Rename from "<avx2_avx512bw>_<shift_insn>v<mode><mask_name>".
(define_insn "<avx2_avx512>_<shift_insn>v<VI2_AVX512VL:mode><mask_name>"):
Rename from "<avx2_avx512bw>_<shift_insn>v<mode><mask_name>".

Co-Authored-By: Andrey Turetskiy <andrey.turetskiy@intel.com>
Co-Authored-By: Anna Tikhonova <anna.tikhonova@intel.com>
Co-Authored-By: Ilya Tocar <ilya.tocar@intel.com>
Co-Authored-By: Ilya Verbin <ilya.verbin@intel.com>
Co-Authored-By: Kirill Yukhin <kirill.yukhin@intel.com>
Co-Authored-By: Maxim Kuznetsov <maxim.kuznetsov@intel.com>
Co-Authored-By: Michael Zolotukhin <michael.v.zolotukhin@intel.com>
From-SVN: r215148

Daily bump.

From-SVN: r215147

vsx.md (vsx_fmav4sf4): Use correct constraints for V2DF, V4SF, DF, and DI modes.

2014-09-10  Michael Meissner  <meissner@linux.vnet.ibm.com>

* config/rs6000/vsx.md (vsx_fmav4sf4): Use correct constraints for
V2DF, V4SF, DF, and DI modes.
(vsx_fmav2df2): Likewise.
(vsx_float_fix_<mode>2): Likewise.
(vsx_reduc_<VEC_reduc_name>_v2df_scalar): Likewise.

From-SVN: r215138

Fix PR target/63209.

From-SVN: r215136

i386.exp: Only run vect-args.c tests if runtest_file_p says they should be run.

* gcc.target/i386/i386.exp: Only run vect-args.c tests
if runtest_file_p says they should be run.

From-SVN: r215135

re PR ipa/61659 (Extra undefined symbol because of devirtualization)

PR c++/61659
* decl.c (grokfndecl): Don't set DECL_COMDAT on static inlines.
(duplicate_decls, start_decl): Likewise.
* pt.c (check_explicit_specialization): Likewise.
(push_template_decl_real): Or static templates.

From-SVN: r215134

error_for_asm and warning_for_asm take const rtx_insn *

gcc/ChangeLog:
2014-09-10  David Malcolm  <dmalcolm@redhat.com>

* final.c (this_is_asm_operands): Strengthen this variable from
rtx to const rtx_insn *.
* output.h (this_is_asm_operands): Likewise.
* rtl-error.c (location_for_asm): Strengthen param "insn" from
const_rtx to const rtx_insn *.
(diagnostic_for_asm): Likewise.
* rtl-error.h (error_for_asm): Likewise.
(warning_for_asm): Likewise.

From-SVN: r215133

insn_extract takes an rtx_insn

gcc/ChangeLog:
2014-09-10  David Malcolm  <dmalcolm@redhat.com>

* genextract.c (print_header): When writing out insn_extract to
insn-extract.c, strengthen the param "insn" from rtx to rtx_insn *.
* recog.h (insn_extract): Strengthen the param from rtx to
rtx_insn *.

From-SVN: r215132

[AArch64 Testsuite] Add a test of the vst[234](q?) intrinics

        * gcc.target/aarch64/vstN_1.c: New test.

From-SVN: r215129

hook.in: Load the xmethods.

2014-09-10  Siva Chandra Reddy  <sivachandra@google.com>

* python/hook.in: Load the xmethods.
* python/Makefile.am (nobase_python_DATA): Add xmethods.py.
* python/Makefile.in: Regenerated.
* python/libstdcxx/v6/xmethods.py: New file.
* testsuite/lib/gdb-test.exp (gdb_version_check_xmethods): New
function.
(gdb-test): New optional argument LOAD_XMETHODS.  Load xmethods
python script if LOAD_XMETHODS is true.
* testsuite/libstdc++-xmethods/unique_ptr.cc: New file.
* testsuite/libstdc++-xmethods/vector.cc: New file.
* testsuite/libstdc++-xmethods/xmethods.exp: New file.

From-SVN: r215128

install.texi (Prerequisites): Note Tcl 8.6 bug fixed in 8.6.1.

* doc/install.texi (Prerequisites): Note Tcl 8.6 bug fixed in
8.6.1.

From-SVN: r215127

[AArch64 Testsuite] Add a test of the vldN_lane intrinsic

        * gcc.target/aarch64/vldN_lane_1.c: New test.

From-SVN: r215126

cgraphunit.c (expand_thunk): If not expanding, set analyzed flag.

2014-09-10  Martin Jambor  <mjambor@suse.cz>

* cgraphunit.c (expand_thunk): If not expanding, set analyzed flag.
(analyze): Do not set analyze flag if expand_thunk returns false;.
(create_wrapper): Likewise.
* cgraphclones.c (duplicate_thunk_for_node): Likewise.

From-SVN: r215123

re PR middle-end/61654 (ICE in release_function_body, at cgraph.c:1699)

2014-09-10  Martin Jambor  <mjambor@suse.cz>

PR ipa/61654
* cgraphclones.c (duplicate_thunk_for_node): Copy arguments of the
new decl properly.  Analyze the new thunk if it is expanded.

gcc/testsuite/
* g++.dg/ipa/pr61654.C: New test.

From-SVN: r215122

coretypes.h (struct _dont_use_rtx_insn_here_, rtx_insn): Define.

* coretypes.h (struct _dont_use_rtx_insn_here_, rtx_insn)
[USED_FOR_TARGET]: Define.

From-SVN: r215121

MIPS: Do not reload unallocated FP_REGS pseudos via GR_REGS

gcc/

* config/mips/mips.c (mips_secondary_reload_class): Handle
regno < 0 case.

From-SVN: r215120

Fix ICE in bitmap routines with LRA and inline assembly language

gcc/

* lra-lives.c (process_bb_lives): Replace assignment with bitwise OR
assignment.

From-SVN: r215119

flag-types.h (enum sanitize_code): Add SANITIZE_NONNULL_ATTRIBUTE and SANITIZE_RETURNS_NONNULL_ATTRIBUTE...

gcc/
* flag-types.h (enum sanitize_code): Add SANITIZE_NONNULL_ATTRIBUTE
and SANITIZE_RETURNS_NONNULL_ATTRIBUTE, or them into SANITIZE_UNDEFINED.
* opts.c (common_handle_option): Handle SANITIZE_NONNULL_ATTRIBUTE and
SANITIZE_RETURNS_NONNULL_ATTRIBUTE and disable
flag_delete_null_pointer_checks for them.
* sanitizer.def (BUILT_IN_UBSAN_HANDLE_NONNULL_ARG,
BUILT_IN_UBSAN_HANDLE_NONNULL_ARG_ABORT,
BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): New.
* ubsan.c (instrument_bool_enum_load): Set *gsi back to
stmt's iterator.
(instrument_nonnull_arg, instrument_nonnull_return): New functions.
(pass_ubsan::gate): Return true even for SANITIZE_NONNULL_ATTRIBUTE
or SANITIZE_RETURNS_NONNULL_ATTRIBUTE.
(pass_ubsan::execute): Call instrument_nonnull_{arg,return}.
* doc/invoke.texi (-fsanitize=nonnull-attribute,
-fsanitize=returns-nonnull-attribute): Document.
gcc/testsuite/
* c-c++-common/ubsan/attrib-3.c: New test.
* c-c++-common/ubsan/nonnull-1.c: New test.
* c-c++-common/ubsan/nonnull-2.c: New test.
* c-c++-common/ubsan/nonnull-3.c: New test.
* c-c++-common/ubsan/nonnull-4.c: New test.
* c-c++-common/ubsan/nonnull-5.c: New test.
libsanitizer/
* ubsan/ubsan_handlers.cc, ubsan/ubsan_handlers.h: Cherry pick
upstream r215485, r217389, r217391 and r217400.

From-SVN: r215118

ubsan.h (struct ubsan_mismatch_data): Removed.

* ubsan.h (struct ubsan_mismatch_data): Removed.
(ubsan_create_data): Remove MISMATCH argument, add LOCCNT argument.
* ubsan.c (ubsan_source_location): For unknown locations,
pass { NULL, 0, 0 } instead of { "<unknown>", x, y }.
(ubsan_create_data): Remove MISMATCH argument, add LOCCNT argument.
Allow more than one location and arbitrary extra arguments passed
in ... instead of through MISMATCH pointer.
(ubsan_instrument_unreachable, ubsan_expand_bounds_ifn,
ubsan_expand_null_ifn, ubsan_build_overflow_builtin,
instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
callers.
c-family/
* c-ubsan.c (ubsan_instrument_division, ubsan_instrument_shift,
ubsan_instrument_vla, ubsan_instrument_return): Adjust
ubsan_create_data callers.
(ubsan_instrument_bounds): Don't emit UBSAN_BOUNDS at all if
index is constant or BIT_AND_EXPR with constant mask and is
small enough for the bound.
* c-gimplify.c (ubsan_walk_array_refs_r): For ADDR_EXPR of
ARRAY_REF, make sure the inner ARRAY_REF is not walked again.

From-SVN: r215117

AVX-512. Add patterns for compress, expand.

gcc/
* config/i386/sse.md
(define_mode_iterator VI48F): New.
(define_insn "<avx512>_compress<mode>_mask"): Rename from
"avx512f_compress<mode>_mask" and update mode iterator.
(define_insn "<avx512>_compressstore<mode>_mask"): Rename from
"avx512f_compressstore<mode>_mask" and update mode iterator.
(define_expand "<avx512>_expand<mode>_maskz"): Rename from
"avx512f_expand<mode>_maskz" and update mode iterator.
(define_insn "<avx512>_expand<mode>_mask"): Rename from
"avx512f_expand<mode>_mask" and update mode iterator.

Co-Authored-By: Andrey Turetskiy <andrey.turetskiy@intel.com>
Co-Authored-By: Anna Tikhonova <anna.tikhonova@intel.com>
Co-Authored-By: Ilya Tocar <ilya.tocar@intel.com>
Co-Authored-By: Ilya Verbin <ilya.verbin@intel.com>
Co-Authored-By: Kirill Yukhin <kirill.yukhin@intel.com>
Co-Authored-By: Maxim Kuznetsov <maxim.kuznetsov@intel.com>
Co-Authored-By: Michael Zolotukhin <michael.v.zolotukhin@intel.com>
From-SVN: r215108

AVX-512. Add reduce, range, fpclass insn patterns.

gcc/
* config/i386/i386.c
(ix86_expand_args_builtin): Handle avx512dq_rangepv8df_mask_round,
avx512dq_rangepv16sf_mask_round, avx512dq_rangepv4df_mask,
avx512dq_rangepv8sf_mask, avx512dq_rangepv2df_mask,
avx512dq_rangepv4sf_mask.
* config/i386/sse.md
(define_c_enum "unspec"): Add UNSPEC_REDUCE, UNSPEC_FPCLASS,
UNSPEC_RANGE.
(define_insn "<mask_codefor>reducep<mode><mask_name>"): New.
(define_insn "reduces<mode>"): Ditto.
(define_insn "avx512dq_rangep<mode><mask_name><round_saeonly_name>"):
Ditto.
(define_insn "avx512dq_ranges<mode><round_saeonly_name>"): Ditto.
(define_insn "avx512dq_fpclass<mode><mask_scalar_merge_name>"): Ditto.
(define_insn "avx512dq_vmfpclass<mode>"): Ditto..

Co-Authored-By: Andrey Turetskiy <andrey.turetskiy@intel.com>
Co-Authored-By: Anna Tikhonova <anna.tikhonova@intel.com>
Co-Authored-By: Ilya Tocar <ilya.tocar@intel.com>
Co-Authored-By: Ilya Verbin <ilya.verbin@intel.com>
Co-Authored-By: Kirill Yukhin <kirill.yukhin@intel.com>
Co-Authored-By: Maxim Kuznetsov <maxim.kuznetsov@intel.com>
Co-Authored-By: Michael Zolotukhin <michael.v.zolotukhin@intel.com>
From-SVN: r215107

AVX-512. Update float unspecs: storeu, rcp14, rsqrt14, scalef, getexp, fixupimm, rndscale, getmant.

gcc/
* config/i386/i386.c
(avx512f_vgetmantv2df_round): Rename from "avx512f_getmantv2df_round".
(avx512f_vgetmantv4sf_round): Rename from "avx512f_vgetmantv4sf_round".
(ix86_expand_args_builtin): Handle avx512vl_getmantv8sf_mask,
avx512vl_getmantv4df_mask, avx512vl_getmantv4sf_mask,
avx512vl_getmantv2df_mask.
(ix86_expand_round_builtin): Handle avx512f_vgetmantv2df_round,
avx512f_vgetmantv4sf_round.
* config/i386/sse.md
(define_insn "<avx512>_storeu<ssemodesuffix><avxsizesuffix>_mask"):
Rename from "avx512f_storeu<ssemodesuffix>512_mask" and update
mode iterator.
(define_insn "<mask_codefor>rcp14<mode><mask_name>"): Use VF_AVX512VL.
(define_insn "<mask_codefor>rsqrt14<mode><mask_name>"): Ditto.
(define_insn "<avx512>_scalef<mode><mask_name><round_name>"): Rename
from "avx512f_scalef<mode><mask_name><round_name>" and update mode
iterator..
(define_insn "<avx512>_getexp<mode><mask_name><round_saeonly_name>"):
Rename from "avx512f_getexp<mode><mask_name><round_saeonly_name>" and
update mode iterator.
(define_expand
"<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>"): Rename from
"avx512f_fixupimm<mode>_maskz<round_saeonly_expand_name>" and update
mode iterator.
(define_insn
"<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>"): Rename
from "avx512f_fixupimm<mode><sd_maskz_name><round_saeonly_name>" and
update mode iterator.
(define_insn "<avx512>_fixupimm<mode>_mask<round_saeonly_name>"): Rename
from "avx512f_fixupimm<mode>_mask<round_saeonly_name>" and update mode
iterator..
(define_insn
"<avx512>_rndscale<mode><mask_name><round_saeonly_name>"): rename from
"avx512f_rndscale<mode><mask_name><round_saeonly_name>" and update
mode iterator..
(define_insn "<avx512>_getmant<mode><mask_name><round_saeonly_name>"):
Rename from "avx512f_getmant<mode><mask_name><round_saeonly_name>" and
update mode iterator.
(define_insn "avx512f_vgetmant<mode><round_saeonly_name>"): Rename from
"avx512f_getmant<mode><round_saeonly_name>".

Co-Authored-By: Andrey Turetskiy <andrey.turetskiy@intel.com>
Co-Authored-By: Anna Tikhonova <anna.tikhonova@intel.com>
Co-Authored-By: Ilya Tocar <ilya.tocar@intel.com>
Co-Authored-By: Ilya Verbin <ilya.verbin@intel.com>
Co-Authored-By: Kirill Yukhin <kirill.yukhin@intel.com>
Co-Authored-By: Maxim Kuznetsov <maxim.kuznetsov@intel.com>
Co-Authored-By: Michael Zolotukhin <michael.v.zolotukhin@intel.com>
From-SVN: r215106

re PR lto/63166 (ICE (LTO): ipa_intraprocedural_devirtualization, at ipa-prop.c:2611)

PR ipa/63166
* ipa-prop.c (compute_known_type_jump_func): Fix conditional.

* g++.dg/lto/pr63166_0.ii: New testcase.
* g++.dg/lto/pr63166_1.ii: New testcase.

From-SVN: r215105

AVX-512. Extend FMA patterns.

gcc/
* config/i386/sse.md (define_mode_iterator VF_AVX512VL): New.
(define_mode_iterator FMAMODEM): Allow 128/256bit EVEX version.
(define_mode_iterator FMAMODE_AVX512): New.
(define_mode_iterator FMAMODE): Remove conditions.
(define_expand "fma4i_fmadd_<mode>"): Use FMAMODE_AVX512 mode iterator.
(define_expand "<avx512>_fmadd_<mode>_maskz<round_expand_name>"): Rename
from "<avx512>_fmadd_<mode>_maskz<round_expand_name>" and use VF_AVX512VL
mode iterator.
(define_mode_iterator FMAMODE_NOVF512): Remove.
(define_insn "*fma_fmadd_<mode>"): Rename from
"<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>" and use
FMAMODE mode iterator.
(define_mode_iterator VF_SF_AVX512VL): New.
(define_insn "<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>"):
Use VF_SF_AVX512VL mode iterator.
(define_insn "<avx512>_fmadd_<mode>_mask<round_name>"): Rename from
"avx512f_fmadd_<mode>_mask<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "<avx512>_fmadd_<mode>_mask3<round_name>"): Rename from
"avx512f_fmadd_<mode>_mask3<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "*fma_fmsub_<mode>"): Rename from
"<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>" and use
FMAMODE mode iterator.
(define_insn "<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>"):
Use VF_SF_AVX512VL mode iterator.
(define_insn "<avx512>_fmsub_<mode>_mask<round_name>"): Rename from
"avx512f_fmsub_<mode>_mask<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "<avx512>_fmsub_<mode>_mask3<round_name>"): Rename from
"avx512f_fmsub_<mode>_mask3<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "*fma_fnmadd_<mode>"): Rename from
"<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>" and
use FMAMODE mode iterator.
(define_insn "<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>"):
Use VF_SF_AVX512VL mode iterator.
(define_insn "<avx512>_fnmadd_<mode>_mask<round_name>"): Rename from
"avx512f_fnmadd_<mode>_mask<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "<avx512>_fnmadd_<mode>_mask3<round_name>"): Rename from
"avx512f_fnmadd_<mode>_mask3<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "*fma_fnmsub_<mode>"): Rename from
"<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>" and use
FMAMODE mode iterator.
(define_insn "<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>"):
Use VF_SF_AVX512VL mode iterator.
(define_insn "<avx512>_fnmsub_<mode>_mask<round_name>"): Rename from
"avx512f_fnmsub_<mode>_mask<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "<avx512>_fnmsub_<mode>_mask3<round_name>"): Rename from
"avx512f_fnmsub_<mode>_mask3<round_name>" and use VF_AVX512VL mode
iterator.
(define_expand "<avx512>_fmaddsub_<mode>_maskz<round_expand_name>"):
Rename from "avx512f_fmaddsub_<mode>_maskz<round_expand_name>" and
use VF_AVX512VL mode iterator.
(define_insn "*fma_fmaddsub_<mode>"): Rename from
"<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>" and
remove subst usage.
(define_insn "<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>"):
Use VF_SF_AVX512VL mode iterator.
(define_insn "<avx512>_fmaddsub_<mode>_mask<round_name>"): Rename from
"avx512f_fmaddsub_<mode>_mask<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "<avx512>_fmaddsub_<mode>_mask3<round_name>"): Rename from
"avx512f_fmaddsub_<mode>_mask3<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "*fma_fmsubadd_<mode>"): Rename from
"<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>" and
remove usage of subst.
(define_insn "<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>"):
Use VF_SF_AVX512VL mode iterator.
(define_insn "<avx512>_fmsubadd_<mode>_mask<round_name>"): Rename from
"avx512f_fmsubadd_<mode>_mask<round_name>" and use VF_AVX512VL mode
iterator.
(define_insn "<avx512>_fmsubadd_<mode>_mask3<round_name>"): Rename from
"avx512f_fmsubadd_<mode>_mask3<round_name>" and use VF_AVX512VL mode
iterator.

Co-Authored-By: Andrey Turetskiy <andrey.turetskiy@intel.com>
Co-Authored-By: Anna Tikhonova <anna.tikhonova@intel.com>
Co-Authored-By: Ilya Tocar <ilya.tocar@intel.com>
Co-Authored-By: Ilya Verbin <ilya.verbin@intel.com>
Co-Authored-By: Kirill Yukhin <kirill.yukhin@intel.com>
Co-Authored-By: Maxim Kuznetsov <maxim.kuznetsov@intel.com>
Co-Authored-By: Michael Zolotukhin <michael.v.zolotukhin@intel.com>
From-SVN: r215104

re PR libgcc/56846 (_Unwind_Backtrace on ARM and noexcept)

2014-09-10  Tony Wang  <tony.wang@arm.com>

    libstdc++-v3/
    PR target/56846
    * libsupc++/eh_personality.cc (PERSONALITY_FUNCTION):
    Return with CONTINUE_UNWINDING when the state pattern
    contains: _US_VIRTUAL_UNWIND_FRAME | _US_FORCE_UNWIND

From-SVN: r215101

revert: calls.c (precompute_arguments): Check promoted_for_signed_and_unsigned_p and set the promoted mode.

gcc/ChangeLog:

2014-09-10  Kugan Vivekanandarajah  <kuganv@linaro.org>

Revert r213751:
* calls.c (precompute_arguments): Check
 promoted_for_signed_and_unsigned_p and set the promoted mode.
(promoted_for_signed_and_unsigned_p): New function.
(expand_expr_real_1): Check promoted_for_signed_and_unsigned_p
and set the promoted mode.
* expr.h (promoted_for_signed_and_unsigned_p): New function definition.
* cfgexpand.c (expand_gimple_stmt_1): Call emit_move_insn if
SUBREG is promoted with SRP_SIGNED_AND_UNSIGNED.

From-SVN: r215100

Daily bump.

From-SVN: r215099

opth-gen.awk: Generate mapping from cpp message reasons to the options that enable them.

gcc/ChangeLog:

2014-09-09  Manuel López-Ibáñez  <manu@gcc.gnu.org>

* opth-gen.awk: Generate mapping from cpp message reasons to the
options that enable them.
* doc/options.texi (CppReason): Document.

gcc/c-family/ChangeLog:

2014-09-09  Manuel López-Ibáñez  <manu@gcc.gnu.org>

* c.opt: Add CppReason to various flags.
(Wdate-time): Re-sort.
* c-common.c: Include c-common.h earlier.
(struct reason_option_codes_t): Delete.
(c_option_controlling_cpp_error): Prefix global type and struct
with cpp_.

From-SVN: r215095

invoke.texi (Wnormalized=): Update.

gcc/ChangeLog:

2014-09-09  Manuel López-Ibáñez  <manu@gcc.gnu.org>

* doc/invoke.texi (Wnormalized=): Update.

libcpp/ChangeLog:

2014-09-09  Manuel López-Ibáñez  <manu@gcc.gnu.org>

* include/cpplib.h (struct cpp_options): Declare warn_normalize as
int instead of enum.

gcc/c-family/ChangeLog:

2014-09-09  Manuel López-Ibáñez  <manu@gcc.gnu.org>

* c.opt (Wnormalized): New.
(Wnormalized=): Use Enum and Reject Negative.
* c-opts.c (c_common_handle_option): Do not handle Wnormalized here.

gcc/testsuite/ChangeLog:

2014-09-09  Manuel López-Ibáñez  <manu@gcc.gnu.org>

* gcc.dg/cpp/warn-normalized-3.c: Delete useless dg-prune-output.

From-SVN: r215093

Fix pad status check.

2014-09-10  Janne Blomqvist  <jb@gcc.gnu.org>

* io/transfer.c (read_block_form): Fix pad status check (found by
Thomas Schwinge with -Wlogical-not-parentheses).

From-SVN: r215092

re PR target/63195 (stage3 build/gengtype miscompiled)

2014-09-09  Segher Boessenkool  <segher@kernel.crashing.org>

PR target/63195
* config/rs6000/rs6000.md (*bool<mode>3): Allow only register
operands.  Split off the constant operand alternative to ...
(*bool<mode>3_imm): New.

From-SVN: r215091

Make std::deque meet C++11 allocator requirements.

* include/bits/deque.tcc (deque::operator=(const deque&)): Handle
allocator propagation.
(deque::emplace_front, deque::emplace_back): Use allocator traits.
(deque::_M_push_back_aux, deque::_M_push_front_aux): Likewise.
(deque::_M_pop_back_aux, deque::_M_pop_front_aux): Likewise.
* include/bits/stl_deque.h (__deque_buf_size): Add constexpr.
(_Deque_iterator): Handle allocators with custom pointers.
(_Deque_base): Likewise. Use allocator traits.
(deque): Likewise. Add allocator-extended constructors.
(deque::_M_move_assign1, deque::_M_move_assign2): Implement move
assignment via tag dispatching.
(deque::_M_replace_map): Replace existing data.
* include/debug/deque (deque): Add allocator-extended constructors.
* include/profile/deque (deque): Likewise.
* testsuite/23_containers/deque/allocator/copy.cc: New.
* testsuite/23_containers/deque/allocator/copy_assign.cc: New.
* testsuite/23_containers/deque/allocator/ext_ptr.cc: New.
* testsuite/23_containers/deque/allocator/minimal.cc: New.
* testsuite/23_containers/deque/allocator/move.cc: New.
* testsuite/23_containers/deque/allocator/move_assign-2.cc: New.
* testsuite/23_containers/deque/allocator/move_assign.cc: New.
* testsuite/23_containers/deque/allocator/noexcept.cc: New.
* testsuite/23_containers/deque/allocator/swap.cc: New.
* testsuite/23_containers/deque/requirements/dr438/assign_neg.cc:
Adjust dg-error line number.
* testsuite/23_containers/deque/requirements/dr438/
constructor_1_neg.cc: Likewise.
* testsuite/23_containers/deque/requirements/dr438/
constructor_2_neg.cc: Likewise.
* testsuite/23_containers/deque/requirements/dr438/insert_neg.cc:
Likewise.
* testsuite/23_containers/vector/52591.cc: Test both the propagating
and always-equal cases.

From-SVN: r215090

single_set takes an insn

gcc/ChangeLog:
2014-09-09  David Malcolm  <dmalcolm@redhat.com>

* rtl.h (single_set_2): Strengthen first param from const_rtx to
const rtx_insn *, and move prototype to above...
(single_set): ...this.  Convert this from a macro to an inline
function, enforcing the requirement that the param is a const
rtx_insn *.
(find_args_size_adjust): Strengthen param from rtx to rtx_insn *.

* config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
Strengthen both params from rtx to rtx_insn *.
* config/arm/aarch-common.c (aarch_crypto_can_dual_issue):
Likewise; introduce locals "producer_set", "consumer_set", using
them in place of "producer" and "consumer" when dealing with SET
rather than insn.
* config/avr/avr.c (avr_out_plus): Add checked cast to rtx_insn *
when invoking single_set in region guarded by INSN_P.
(avr_out_bitop): Likewise.
(_reg_unused_after): Introduce local rtx_sequence * "seq" in
region guarded by GET_CODE check, using methods to strengthen
local "this_insn" from rtx to rtx_insn *, and for clarity.
* config/avr/avr.md (define_insn_and_split "xload8<mode>_A"):
Strengthen local "insn" from rtx to rtx_insn *.
(define_insn_and_split "xload<mode>_A"): Likewise.
* config/bfin/bfin.c (trapping_loads_p): Likewise for param
"insn".
(find_load): Likewise for return type.
(workaround_speculation): Likewise for both locals named
"load_insn".
* config/cris/cris.c (cris_cc0_user_requires_cmp): Likewise for
local "cc0_user".
* config/cris/cris.md (define_peephole2 ; moversideqi): Likewise
for local "prev".
* config/h8300/h8300-protos.h (notice_update_cc): Likewise for
param 2.
* config/h8300/h8300.c (notice_update_cc): Likewise.
* config/i386/i386.c (ix86_flags_dependent): Likewise for params
"insn" and "dep_insn".
(exact_store_load_dependency): Likewise for both params.
(ix86_macro_fusion_pair_p): Eliminate local named "single_set"
since this now clashes with inline function.  Instead, delay
calling single_set until the point where its needed, and then
assign the result to "compare_set" and rework the conditional that
follows.
* config/ia64/ia64.md (define_expand "tablejump"): Strengthen
local "last" from rtx to rtx_insn *.
* config/mips/mips-protos.h (mips_load_store_insns): Likewise for
second param.
(mips_store_data_bypass_p): Likewise for both params.
* config/mips/mips.c (mips_load_store_insns): Likewise for second
param.
(mips_store_data_bypass_p): Likewise for both params.
(mips_orphaned_high_part_p): Likewise for param "insn".
* config/mn10300/mn10300.c (extract_bundle): Likewise.
(mn10300_bundle_liw): Likewise for locals "r", "insn1", "insn2".
Introduce local rtx "insn2_pat".
* config/rl78/rl78.c (move_elim_pass): Likewise for locals "insn",
"ninsn".
(rl78_remove_unused_sets): Likewise for locals "insn", "ninsn".
Introduce local rtx "set", using it in place of "insn" for the
result of single_set.  This appears to fix a bug, since the call
to find_regno_note on a SET does nothing.
* config/rs6000/rs6000.c (set_to_load_agen): Strengthen both
params from rtx to rtx_insn *.
(set_to_load_agen): Likewise.
* config/s390/s390.c (s390_label_align): Likewise for local
"prev_insn".  Introduce new rtx locals "set" and "src", using
them in place of "prev_insn" for the results of single_set
and SET_SRC respectively.
(s390_swap_cmp): Strengthen local "jump" from rtx to rtx_insn *.
Introduce new rtx local "set" using in place of "jump" for the
result of single_set.  Use SET_SRC (set) rather than plain
XEXP (set, 1).
* config/sh/sh.c (noncall_uses_reg): Strengthen param 2from
rtx to rtx_insn *.
(noncall_uses_reg): Likewise.
(reg_unused_after): Introduce local rtx_sequence * "seq" in region
guarded by GET_CODE check, using its methods for clarity, and to
enable strengthening local "this_insn" from rtx to rtx_insn *.
* config/sh/sh.md (define_expand "mulhisi3"): Strengthen local
"insn" from rtx to rtx_insn *.
(define_expand "umulhisi3"): Likewise.
(define_expand "smulsi3_highpart"): Likewise.
(define_expand "umulsi3_highpart"): Likewise.
* config/sparc/sparc.c (sparc_do_work_around_errata): Likewise for
local "after".  Replace GET_CODE check with a dyn_cast,
introducing new local rtx_sequence * "seq", using insn method for
typesafety.

* dwarf2cfi.c (dwarf2out_frame_debug): Strengthen param "insn"
from rtx to rtx_insn *.  Introduce local rtx "pat", using it in
place of "insn" once we're dealing with patterns rather than the
input insn.
(scan_insn_after): Strengthen param "insn" from rtx to rtx_insn *.
(scan_trace): Likewise for local "elt", updating lookups within
sequence to use insn method rather than element method.
* expr.c (find_args_size_adjust): Strengthen param "insn" from rtx
to rtx_insn *.
* gcse.c (gcse_emit_move_after): Likewise for local "new_rtx".
* ifcvt.c (noce_try_abs): Likewise for local "insn".
* ira.c (fix_reg_equiv_init): Add checked cast to rtx_insn * when
invoking single_set.
* lra-constraints.c (insn_rhs_dead_pseudo_p): Strengthen param
"insn" from rtx to rtx_insn *.
(skip_usage_debug_insns): Likewise for return type, adding a
checked cast.
(check_secondary_memory_needed_p): Likewise for local "insn".
(inherit_reload_reg): Likewise.
* modulo-sched.c (sms_schedule): Likewise for local "count_init".
* recog.c (peep2_attempt): Likewise for local "old_insn", adding
checked casts.
(store_data_bypass_p): Likewise for both params.
(if_test_bypass_p): Likewise.
* recog.h (store_data_bypass_p): Likewise for both params.
(if_test_bypass_p): Likewise.
* reload.c (find_equiv_reg): Likewise for local "where".
* reorg.c (delete_jump): Likewise for param "insn".
* rtlanal.c (single_set_2): Strenghen param "insn" from const_rtx
to const rtx_insn *.
* store-motion.c (replace_store_insn): Likewise for param "del".
(delete_store): Strengthen local "i" from rtx to rtx_insn_list *,
and use its methods for clarity, and to strengthen local "del"
from rtx to rtx_insn *.
(build_store_vectors): Use insn method of "st" when calling
replace_store_insn for typesafety and clarity.

From-SVN: r215089

rs6000.c (rtx_is_swappable_p): Add UNSPEC_VSX_CVDPSPN as an unswappable operand...

[gcc]

2014-09-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/rs6000.c (rtx_is_swappable_p): Add
UNSPEC_VSX_CVDPSPN as an unswappable operand, and add commentary
on how to make it legal in future.

[gcc/testsuite]

2014-09-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gcc.target/powerpc/swaps-p8-15.c: Remove scan-assembler-not for
xxpermdi and add commentary about adding it back later; remove
unused typedef.

From-SVN: r215088

recog_memoized works on an rtx_insn *

gcc/ChangeLog:
2014-09-09  David Malcolm  <dmalcolm@redhat.com>

* caller-save.c (rtx saveinsn): Strengthen this variable from rtx
to rtx_insn *.
(restinsn): Likewise.
* config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_move):
Likewise for param.
* config/aarch64/aarch64.c (aarch64_simd_attr_length_move):
Likewise.
* config/arc/arc-protos.h (arc_adjust_insn_length): Likewise for
first param.
(arc_hazard): Likewise for both params.
* config/arc/arc.c (arc600_corereg_hazard): Likewise, adding
checked casts to rtx_sequence * and uses of the insn method for
type-safety.
(arc_hazard): Strengthen both params from rtx to rtx_insn *.
(arc_adjust_insn_length): Likewise for param "insn".
(struct insn_length_parameters_s): Likewise for first param of
"get_variants" callback field.
(arc_get_insn_variants): Likewise for first param and local
"inner".  Replace a check of GET_CODE with a dyn_cast to
rtx_sequence *, using methods for type-safety and clarity.
* config/arc/arc.h (ADJUST_INSN_LENGTH): Use casts to
rtx_sequence * and uses of the insn method for type-safety when
invoking arc_adjust_insn_length.
* config/arm/arm-protos.h (arm_attr_length_move_neon): Likewise
for param.
(arm_address_offset_is_imm): Likewise.
(struct tune_params): Likewise for params 1 and 3 of the
"sched_adjust_cost" callback field.
* config/arm/arm.c (cortex_a9_sched_adjust_cost): Likewise for
params 1 and 3 ("insn" and "dep").
(xscale_sched_adjust_cost): Likewise.
(fa726te_sched_adjust_cost): Likewise.
(cortexa7_older_only): Likewise for param "insn".
(cortexa7_younger): Likewise.
(arm_attr_length_move_neon): Likewise.
(arm_address_offset_is_imm): Likewise.
* config/avr/avr-protos.h (avr_notice_update_cc): Likewise.
* config/avr/avr.c (avr_notice_update_cc): Likewise.
* config/bfin/bfin.c (hwloop_pattern_reg): Likewise.
(workaround_speculation): Likewise for local "last_condjump".
* config/c6x/c6x.c (shadow_p): Likewise for param "insn".
(shadow_or_blockage_p): Likewise.
(get_unit_reqs): Likewise.
(get_unit_operand_masks): Likewise.
(c6x_registers_update): Likewise.
(returning_call_p): Likewise.
(can_use_callp): Likewise.
(convert_to_callp): Likewise.
(find_last_same_clock): Likwise for local "t".
(reorg_split_calls): Likewise for local "shadow".
(hwloop_pattern_reg): Likewise for param "insn".
* config/frv/frv-protos.h (frv_final_prescan_insn): Likewise.
* config/frv/frv.c (frv_final_prescan_insn): Likewise.
(frv_extract_membar): Likewise.
(frv_optimize_membar_local): Strengthen param "last_membar" from
rtx * to rtx_insn **.
(frv_optimize_membar_global): Strengthen param "membar" from rtx
to rtx_insn *.
(frv_optimize_membar): Strengthen local "last_membar" from rtx *
to rtx_insn **.
* config/ia64/ia64-protos.h (ia64_st_address_bypass_p): Strengthen
both params from rtx to rtx_insn *.
(ia64_ld_address_bypass_p): Likewise.
* config/ia64/ia64.c (ia64_safe_itanium_class): Likewise for param
"insn".
(ia64_safe_type): Likewise.
(group_barrier_needed): Likewise.
(safe_group_barrier_needed): Likewise.
(ia64_single_set): Likewise.
(is_load_p): Likewise.
(record_memory_reference): Likewise.
(get_mode_no_for_insn): Likewise.
(important_for_bundling_p): Likewise.
(unknown_for_bundling_p): Likewise.
(ia64_st_address_bypass_p): Likewise for both params.
(ia64_ld_address_bypass_p): Likewise.
(expand_vselect): Introduce new local rtx_insn * "insn", using it
in place of rtx "x" after the emit_insn call.
* config/i386/i386-protos.h (x86_extended_QIreg_mentioned_p):
Strengthen param from rtx to rtx_insn *.
(ix86_agi_dependent): Likewise for both params.
(ix86_attr_length_immediate_default): Likewise for param 1.
(ix86_attr_length_address_default): Likewise for param.
(ix86_attr_length_vex_default): Likewise for param 1.
* config/i386/i386.c (ix86_attr_length_immediate_default):
Likewise for param "insn".
(ix86_attr_length_address_default): Likewise.
(ix86_attr_length_vex_default): Likewise.
(ix86_agi_dependent): Likewise for both params.
(x86_extended_QIreg_mentioned_p): Likewise for param "insn".
(vselect_insn): Likewise for this variable.
* config/m68k/m68k-protos.h (m68k_sched_attr_opx_type): Likewise
for param 1.
(m68k_sched_attr_opy_type): Likewise.
* config/m68k/m68k.c (sched_get_operand): Likewise.
(sched_attr_op_type): Likewise.
(m68k_sched_attr_opx_type): Likewise.
(m68k_sched_attr_opy_type): Likewise.
(sched_get_reg_operand): Likewise.
(sched_get_mem_operand): Likewise.
(m68k_sched_address_bypass_p): Likewise for both params.
(sched_get_indexed_address_scale): Likewise.
(m68k_sched_indexed_address_bypass_p): Likewise.
* config/m68k/m68k.h (m68k_sched_address_bypass_p): Likewise.
(m68k_sched_indexed_address_bypass_p): Likewise.
* config/mep/mep.c (mep_jmp_return_reorg): Strengthen locals
"label", "ret" from rtx to rtx_insn *, adding a checked cast and
removing another.
* config/mips/mips-protos.h (mips_linked_madd_p): Strengthen both
params from rtx to rtx_insn *.
(mips_fmadd_bypass): Likewise.
* config/mips/mips.c (mips_fmadd_bypass): Likewise.
(mips_linked_madd_p): Likewise.
(mips_macc_chains_last_hilo): Likewise for this variable.
(mips_macc_chains_record): Likewise for param.
(vr4130_last_insn): Likewise for this variable.
(vr4130_swap_insns_p): Likewise for both params.
(mips_ls2_variable_issue): Likewise for param.
(mips_need_noat_wrapper_p): Likewise for param "insn".
(mips_expand_vselect): Add a new local rtx_insn * "insn", using it
in place of "x" after the emit_insn.
* config/pa/pa-protos.h (pa_fpstore_bypass_p): Strengthen both
params from rtx to rtx_insn *.
* config/pa/pa.c (pa_fpstore_bypass_p): Likewise.
(pa_combine_instructions): Introduce local "par" for result of
gen_rtx_PARALLEL, moving decl and usage of new_rtx for after call
to make_insn_raw.
(pa_can_combine_p): Strengthen param "new_rtx" from rtx to rtx_insn *.
* config/rl78/rl78.c (insn_ok_now): Likewise for param "insn".
(rl78_alloc_physical_registers_op1): Likewise.
(rl78_alloc_physical_registers_op2): Likewise.
(rl78_alloc_physical_registers_ro1): Likewise.
(rl78_alloc_physical_registers_cmp): Likewise.
(rl78_alloc_physical_registers_umul): Likewise.
(rl78_alloc_address_registers_macax): Likewise.
(rl78_alloc_physical_registers): Likewise for locals "insn", "curr".
* config/s390/predicates.md (execute_operation): Likewise for
local "insn".
* config/s390/s390-protos.h (s390_agen_dep_p): Likewise for both
params.
* config/s390/s390.c (s390_safe_attr_type): Likewise for param.
(addr_generation_dependency_p): Likewise for param "insn".
(s390_agen_dep_p): Likewise for both params.
(s390_fpload_toreg): Likewise for param "insn".
* config/sh/sh-protos.h (sh_loop_align): Likewise for param.
* config/sh/sh.c (sh_loop_align): Likewise for param and local
"next".
* config/sh/sh.md (define_peephole2): Likewise for local "insn2".
* config/sh/sh_treg_combine.cc
(sh_treg_combine::make_inv_ccreg_insn): Likewise for return type
and local "i".
(sh_treg_combine::try_eliminate_cstores): Likewise for local "i".
* config/stormy16/stormy16.c (combine_bnp): Likewise for locals
"and_insn", "load", "shift".
* config/tilegx/tilegx.c (match_pcrel_step2): Likewise for param
"insn".
* final.c (final_scan_insn): Introduce local rtx_insn * "other"
for XEXP (note, 0) of the REG_CC_SETTER note.
(cleanup_subreg_operands): Strengthen param "insn" from rtx to
rtx_insn *, eliminating a checked cast made redundant by this.
* gcse.c (process_insert_insn): Strengthen local "insn" from rtx
to rtx_insn *.
* genattr.c (main): When writing out the prototype to
const_num_delay_slots, strengthen the param from rtx to
rtx_insn *.
* genattrtab.c (write_const_num_delay_slots): Likewise when
writing out the implementation of const_num_delay_slots.
* hw-doloop.h (struct hw_doloop_hooks): Strengthen the param
"insn" of callback field "end_pattern_reg" from rtx to rtx_insn *.
* ifcvt.c (noce_emit_store_flag): Eliminate local rtx "tmp" in
favor of new rtx locals "src" and "set" and new local rtx_insn *
"insn" and "seq".
(noce_emit_move_insn): Strengthen locals "seq" and "insn" from rtx
to rtx_insn *.
(noce_emit_cmove): Eliminate local rtx "tmp" in favor of new rtx
locals "cond", "if_then_else", "set" and new rtx_insn * locals
"insn" and "seq".
(noce_try_cmove_arith): Strengthen locals "insn_a" and "insn_b",
"last" from rtx to rtx_insn *.  Likewise for a local "tmp",
renaming to "tmp_insn".  Eliminate the other local rtx "tmp" from
the top-level scope, replacing with new more tightly-scoped rtx
locals "reg", "pat", "mem" and rtx_insn * "insn", "copy_of_a",
"new_insn", "copy_of_insn_b", and make local rtx "set" more
tightly-scoped.
* ira-int.h (ira_setup_alts): Strengthen param "insn" from rtx to
rtx_insn *.
* ira.c (setup_prohibited_mode_move_regs): Likewise for local
"move_insn".
(ira_setup_alts): Likewise for param "insn".
* lra-constraints.c (emit_inc): Likewise for local "add_insn".
* lra.c (emit_add3_insn): Split local rtx "insn" in two, an rtx
and an rtx_insn *.
(lra_emit_add): Eliminate top-level local rtx "insn" in favor of
new more-tightly scoped rtx locals "add3_insn", "insn",
"add2_insn" and rtx_insn * "move_insn".
* postreload-gcse.c (eliminate_partially_redundant_load): Add
checked cast on result of gen_move_insn when invoking
extract_insn.
* recog.c (insn_invalid_p): Strengthen param "insn" from rtx to
rtx_insn *.
(verify_changes): Add a checked cast on "object" when invoking
insn_invalid_p.
(extract_insn_cached): Strengthen param "insn" from rtx to
rtx_insn *.
(extract_constrain_insn_cached): Likewise.
(extract_insn): Likewise.
* recog.h (insn_invalid_p): Likewise for param 1.
(recog_memoized): Likewise for param.
(extract_insn): Likewise.
(extract_constrain_insn_cached): Likewise.
(extract_insn_cached): Likewise.
* reload.c (can_reload_into): Likewise for local "test_insn".
* reload.h (cleanup_subreg_operands): Likewise for param.
* reload1.c (emit_insn_if_valid_for_reload): Rename param from
"insn" to "pat", reintroducing "insn" as an rtx_insn * on the
result of emit_insn.  Remove a checked cast made redundant by this
change.
* sel-sched-ir.c (sel_insn_rtx_cost): Strengthen param "insn" from
rtx to rtx_insn *.
* sel-sched.c (get_reg_class): Likewise.

From-SVN: r215087

Add crtfastmath for AArch64.

gcc/Changelog

2014-09-09  Marcus Shawcroft  <marcus.shawcroft@arm.com>
            Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>

     * config/aarch64/aarch64-elf-raw.h (ENDFILE_SPEC): Add crtfastmath.o.
     * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_MATH_ENDFILE_SPEC): Define.
      (ENDFILE_SPEC): Define and use GNU_USER_TARGET_MATH_ENDFILE_SPEC.

libgcc/Changelog

2014-09-09  Marcus Shawcroft  <marcus.shawcroft@arm.com>
            Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>

    * config.host (aarch64*): Include crtfastmath and t-crtfm.
    * config/aarch64/crtfastmath.c: New file.

Co-Authored-By: Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
From-SVN: r215086

[ARM][tests] Make input and output arrays 128-bit aligned in vectorisation tests

* gcc.target/arm/vect-lceilf_1.c: Make input and output arrays global
and 16-byte aligned.
* gcc.target/arm/vect-lfloorf_1.c: Likewise.
* gcc.target/arm/vect-lroundf_1.c: Likewise.
* gcc.target/arm/vect-rounding-btruncf.c: Likewise.
* gcc.target/arm/vect-rounding-ceilf.c: Likewise.
* gcc.target/arm/vect-rounding-floorf.c: Likewise.
* gcc.target/arm/vect-rounding-roundf.c: Likewise.

From-SVN: r215085

INSN_LOCATION takes an rtx_insn

gcc/ChangeLog:
2014-09-09  David Malcolm  <dmalcolm@redhat.com>

* rtl.h (INSN_LOCATION): Strengthen param from const_rtx to
const rtx_insn *, and from rtx to rtx_insn * for the other
overloaded variant.
(RTL_LOCATION): Add a checked cast to rtx_insn * when invoking
INSN_LOCATION, since we know INSN_P holds.
(insn_line): Strengthen param from const_rtx to const rtx_insn *.
(insn_file): Likewise.
(insn_scope): Likewise.
(insn_location): Likewise.

* config/mips/mips.c (mips16_gp_pseudo_reg): Strengthen local
"insn" from rtx to rtx_insn *, introducing a new local rtx "set"
for the result of gen_load_const_gp.
* config/rs6000/rs6000-protos.h (output_call): Strengthen first
param from rtx to rtx_insn *.
* config/rs6000/rs6000.c (output_call): Likewise.
* dwarf2out.c (dwarf2out_var_location): Likewise for local "prev",
introducing a checked cast to rtx_sequence * and use of the insn
method.
* emit-rtl.c (emit_copy_of_insn_after): Strengthen both params
from rtx to rtx_insn *.
(insn_scope): Strengthen param from const_rtx to const rtx_insn *.
(insn_line): Likewise.
(insn_file): Likewise.
(insn_location): Likewise.
* emit-rtl.h (emit_copy_of_insn_after): Strengthen both params
from rtx to rtx_insn *.
* print-rtl.c (print_rtx): Introduce local "in_insn" via a checked
cast, using it for calls to INSN_HAS_LOCATION and insn_location.
* recog.c (peep2_attempt): Introduce local rtx_insn * "peepinsn"
via a checked cast.
* reorg.c (relax_delay_slots): Strengthen locals named "after"
from rtx to rtx_insn *; use methods of "pat" for type-safety.

From-SVN: r215084

Handcode gen_rtx_INSN

gcc/ChangeLog:
2014-09-09  David Malcolm  <dmalcolm@redhat.com>

* combine.c (try_combine): Eliminate checked cast on result of
gen_rtx_INSN.
* emit-rtl.c (gen_rtx_INSN): New function, improving over the prior
autogenerated one by strengthening the return type and params 2 and 3
from rtx to rtx_insn *, and by naming the params.
* gengenrtl.c (special_rtx): Add INSN to those that are
special-cased.
* rtl.h (gen_rtx_INSN): New prototype.

From-SVN: r215083

struct ira_reg_equiv_s's "init_insns" is an rtx_insn_list

gcc/ChangeLog:
2014-09-09  David Malcolm  <dmalcolm@redhat.com>

* ira.c (ira_update_equiv_info_by_shuffle_insn): Use NULL rather
than NULL_RTX.
(no_equiv): Likewise.
(update_equiv_regs): Likewise.
(setup_reg_equiv): Likewise.  Strengthen locals "elem",
"prev_elem", "next_elem" from rtx to rtx_insn_list *, and "insn"
from rtx to rtx_insn *.  Use methods of "elem" for typesafety and
clarity.
* ira.h (struct ira_reg_equiv_s): Strengthen field "init_insns"
from rtx to rtx_insn_list *.
* lra-assigns.c (spill_for): Strengthen local "x" from rtx to
rtx_insn_list * and use methods for clarity and typesafety.
* lra-constraints.c (contains_deleted_insn_p): Likewise for param
"list".
(init_insn_rhs_dead_pseudo_p): Likewise for local "insns".  Remove
redundant check on INSN_P (insns): this cannot hold, as "insns" is
an INSN_LIST, not an insn.
(reverse_equiv_p): Strengthen local "insns" from rtx to
rtx_insn_list * and use methods for clarity and typesafety.
(contains_reloaded_insn_p): Likewise for local "list".

From-SVN: r215082

[AArch64 Testsuite] Add a test of vldN_dup intrinsics

        * gcc.target/aarch64/vldN_dup_1.c: New test.

From-SVN: r215078

[AArch64 Testsuite] Extend test of vld1+vst1 intrinsics to cover more variants

* gcc.target/aarch64/vld1-vst1_1.c: Rewrite to test all variants.

From-SVN: r215077

Update email address for recent commits, to comply with company policy.

From-SVN: r215074

[AArch64 Testsuite] Add test of vld[234]q? intrinsic

* gcc.target/aarch64/vldN_1.c: New test.

From-SVN: r215072

[AArch64 Testsuite]Fix scan-assembler test false alarm on aarch64-linux-gnu

        * gcc.target/aarch64/simd/int_comparisons_1.c: Tighten regexp.

From-SVN: r215071

typeck.c (build_class_member_access_expr): Move -Winvalid-offsetof code...

* typeck.c (build_class_member_access_expr): Move
-Winvalid-offsetof code...
* semantics.c (finish_offsetof): ...here.
* parser.c (cp_parser_builtin_offsetof): Remember the location of
the type argument.
* pt.c (tsubst_copy_and_build) [OFFSETOF_EXPR]: Preserve it.

From-SVN: r215070

[ARM] Enable auto-vectorization for copysignf

  gcc/
    * config/arm/arm.c (NEON_COPYSIGNF): New enum.
    (arm_init_neon_builtins): Support NEON_COPYSIGNF.
    (arm_builtin_vectorized_function): Likewise.
    * config/arm/arm_neon_builtins.def: New macro for copysignf.
    * config/arm/neon.md (neon_copysignf<mode>): New pattern for vector
    copysignf.

  gcc/testsuite/
    * gcc.target/arm/vect-copysignf.c: New testcase.

From-SVN: r215067

bb-reorder.h (default_target_bb_reorder): Remove redundant GTY.

gcc/
* bb-reorder.h (default_target_bb_reorder): Remove redundant GTY.
* builtins.h (default_target_builtins): Likewise.
* gcse.h (default_target_gcse): Likewise.
* target-globals.h (target_globals): Add a destructor.  Convert
void-pointer fields back to their real type and change from
GTY((atomic)) to GTY((skip)).
(restore_target_globals): Remove casts accordingly.
* target-globals.c (save_target_globals): Use XCNEW rather than
ggc_internal_cleared_alloc to allocate non-GC structures.
Use ggc_cleared_alloc to allocate the target_globals structure
itself.
(target_globals::~target_globals): Define.

From-SVN: r215063

re PR c++/62255 (Introducing an unrelated template parameter causes compilation to fail)

PR c++/62255
* pt.c (instantiate_decl): Handle recursive instantiation of
static data member.

From-SVN: r215062

[ARM][7/7] Convert FP mnemonics to UAL | f{ld,st}m -> v{ld,st}m

* config/arm/arm.md (vfp_pop_multiple_with_writeback): Use vldm
mnemonic instead of fldmfdd.
* config/arm/arm.c (vfp_output_fstmd): Rename to...
(vfp_output_vstmd): ... This.  Convert output to UAL syntax.
Output vpush when address register is SP.
* config/arm/arm-protos.h (vfp_output_fstmd): Rename to...
(vfp_output_vstmd): ... This.
* config/arm/vfp.md (push_multi_vfp): Update call to
vfp_output_vstmd.

From-SVN: r215056

[ARM][6/7] Convert FP mnemonics to UAL | movcc_vfp (fmstat)

* config/arm/vfp.md (*movcc_vfp): Use UAL syntax.

From-SVN: r215055

[ARM][5/7] Convert FP mnemonics to UAL | sqrt and FP compare patterns

* config/arm/vfp.md (*sqrtsf2_vfp): Use UAL assembly syntax.
(*sqrtdf2_vfp): Likewise.
(*cmpsf_vfp): Likewise.
(*cmpsf_trap_vfp): Likewise.
(*cmpdf_vfp): Likewise.
(*cmpdf_trap_vfp): Likewise.

* gcc.target/arm/vfp-1.c: Updated expected assembly.

From-SVN: r215054

[ARM][4/7] Convert FP mnemonics to UAL | vcvt patterns

* config/arm/vfp.md (*extendsfdf2_vfp): Use UAL assembly syntax.
(*truncdfsf2_vfp): Likewise.
(*truncsisf2_vfp): Likewise.
(*truncsidf2_vfp): Likewise.
(fixuns_truncsfsi2): Likewise.
(fixuns_truncdfsi2): Likewise.
(*floatsisf2_vfp): Likewise.
(*floatsidf2_vfp): Likewise.
(floatunssisf2): Likewise.
(floatunssidf2): Likewise.

* gcc.target/arm/vfp-1.c: Updated expected assembly.

From-SVN: r215053

[3/7] Convert FP mnemonics to UAL | mul+add patterns

* config/arm/vfp.md (*mulsf3_vfp): Use UAL assembly syntax.
(*muldf3_vfp): Likewise.
(*mulsf3negsf_vfp): Likewise.
(*muldf3negdf_vfp): Likewise.
(*mulsf3addsf_vfp): Likewise.
(*muldf3adddf_vfp): Likewise.
(*mulsf3subsf_vfp): Likewise.
(*muldf3subdf_vfp): Likewise.
(*mulsf3negsfaddsf_vfp): Likewise.
(*fmuldf3negdfadddf_vfp): Likewise.
(*mulsf3negsfsubsf_vfp): Likewise.
(*muldf3negdfsubdf_vfp): Likewise.

* gcc.target/arm/vfp-1.c: Updated expected assembly.

From-SVN: r215052

[ARM][2/7] Convert FP mnemonics to UAL | add/sub/div/abs patterns.

* config/arm/vfp.md (*abssf2_vfp): Use UAL assembly syntax.
(*absdf2_vfp): Likewise.
(*negsf2_vfp): Likewise.
(*negdf2_vfp): Likewise.
(*addsf3_vfp): Likewise.
(*adddf3_vfp): Likewise.
(*subsf3_vfp): Likewise.
(*subdf3_vfp): Likewise.
(*divsf3_vfp): Likewise.
(*divdf3_vfp): Likewise.

* gcc.target/arm/vfp-1.c: Updated expected assembly.

From-SVN: r215051

[ARM][1/7] Convert FP mnemonics to UAL | mov patterns.

* config/arm/arm.c (output_move_vfp): Use UAL syntax for load/store
multiple.
(arm_print_operand): Don't convert real values to decimal
representation in default case.
(fp_immediate_constant): Delete.
* config/arm/arm-protos.h (fp_immediate_constant): Likewise.
* config/arm/vfp.md (*arm_movsi_vfp): Convert to VFP moves to UAL
syntax.
(*thumb2_movsi_vfp): Likewise.
(*movdi_vfp): Likewise.
(*movdi_vfp_cortexa8): Likewise.
(*movhf_vfp_neon): Likewise.
(*movhf_vfp): Likewise.
(*movsf_vfp): Likewise.
(*thumb2_movsf_vfp): Likewise.
(*movdf_vfp): Likewise.
(*thumb2_movdf_vfp): Likewise.
(*movsfcc_vfp): Likewise.
(*thumb2_movsfcc_vfp): Likewise.
(*movdfcc_vfp): Likewise.
(*thumb2_movdfcc_vfp): Likewise.

* gcc.target/arm/pr51835.c: Update expected assembly.
* gcc.target/arm/vfp-1.c: Likewise.
* gcc.target/arm/vfp-ldmdbd.c: Likewise.
* gcc.target/arm/vfp-ldmdbs.c: Likewise.
* gcc.target/arm/vfp-ldmiad.c: Likewise.
* gcc.target/arm/vfp-ldmias.c: Likewise.
* gcc.target/arm/vfp-stmdbd.c: Likewise.
* gcc.target/arm/vfp-stmdbs.c: Likewise.
* gcc.target/arm/vfp-stmiad.c: Likewise.
* gcc.target/arm/vfp-stmias.c: Likewise.

From-SVN: r215050

[Obvious] Change Linux to GNU/LINUX in ARM options documentation

gcc/
* doc/invoke.texi (-march): Use GNU/Linux rather than Linux.
(-mtune): Likewise.
(-mcpu): Likewise.

From-SVN: r215049

xordi3-opt.c: Disable this test case for thumb1 target.

2014-09-09  Tony Wang  <tony.wang@arm.com>

    gcc/testsuite/
    * gcc.target/arm/xordi3-opt.c: Disable this
    test case for thumb1 target.
    * gcc.target/arm/iordi3-opt.c: Ditto.

From-SVN: r215047

[AArch64] PR 61749: Do not ICE in lane intrinsics when passed non-constant lane number

PR target/61749
* config/aarch64/aarch64-builtins.c (aarch64_types_quadop_qualifiers):
Use qualifier_immediate for last operand.  Rename to...
(aarch64_types_ternop_lane_qualifiers): ... This.
(TYPES_QUADOP): Rename to...
(TYPES_TERNOP_LANE): ... This.
(aarch64_simd_expand_args): Return const0_rtx when encountering user
error.  Change return of 0 to return of NULL_RTX.
(aarch64_crc32_expand_builtin): Likewise.
(aarch64_expand_builtin): Return NULL_RTX instead of 0.
ICE when expanding unknown builtin.
* config/aarch64/aarch64-simd-builtins.def (sqdmlal_lane): Use
TERNOP_LANE qualifiers.
(sqdmlsl_lane): Likewise.
(sqdmlal_laneq): Likewise.
(sqdmlsl_laneq): Likewise.
(sqdmlal2_lane): Likewise.
(sqdmlsl2_lane): Likewise.
(sqdmlal2_laneq): Likewise.
(sqdmlsl2_laneq): Likewise.

* gcc.target/aarch64/vqdml_lane_intrinsics-bad_1.c: New test.

From-SVN: r215046

invoke.texi (Optimization Options): Add missing @gol to the end of a line.

* doc/invoke.texi (Optimization Options): Add missing @gol to the
end of a line.
(S/390 and zSeries Options): Remove superfluous word from the
description of the -mhotpatch option.

From-SVN: r215044

shrink-wrap.h: #define SHRINK_WRAPPING_ENABLED.

ChangeLog:
2014-09-09  Zhenqiang Chen  <zhenqiang.chen@arm.com>

* shrink-wrap.h: #define SHRINK_WRAPPING_ENABLED.
* ira.c: #include "shrink-wrap.h"
(split_live_ranges_for_shrink_wrap): Use SHRINK_WRAPPING_ENABLED.
* ifcvt.c: #include "shrink-wrap.h"
(dead_or_predicable): Use SHRINK_WRAPPING_ENABLED.

testsuite/ChangeLog:
2014-09-09  Zhenqiang Chen  <zhenqiang.chen@arm.com>

* gcc.target/arm/split-live-ranges-for-shrink-wrap.c: New test.

From-SVN: r215041

remove picochip

contrib/ChangeLog:

2014-09-08  Trevor Saunders  <tsaunders@mozilla.com>

* compare-all-tests: Don't test picochip.
* config-list.mk: Likewise.

gcc/ChangeLog:

2014-09-08  Trevor Saunders  <tsaunders@mozilla.com>

* common/config/picochip/picochip-common.c: Remove.
* config.gcc: Remove support for picochip.
* config/picochip/constraints.md: Remove.
* config/picochip/dfa_space.md: Remove.
* config/picochip/dfa_speed.md: Remove.
* config/picochip/picochip-protos.h: Remove.
* config/picochip/picochip.c: Remove.
* config/picochip/picochip.h: Remove.
* config/picochip/picochip.md: Remove.
* config/picochip/picochip.opt: Remove.
* config/picochip/predicates.md: Remove.
* config/picochip/t-picochip: Remove.
* doc/md.texi: Don't document picochi.

libgcc/ChangeLog:

2014-09-08  Trevor Saunders  <tsaunders@mozilla.com>

* config.host: Remove picochip support.
* config/picochip/adddi3.S: Remove.
* config/picochip/ashlsi3.S: Remove.
* config/picochip/ashlsi3.c: Remove.
* config/picochip/ashrsi3.S: Remove.
* config/picochip/ashrsi3.c: Remove.
* config/picochip/clzsi2.S: Remove.
* config/picochip/cmpsi2.S: Remove.
* config/picochip/divmod15.S: Remove.
* config/picochip/divmodhi4.S: Remove.
* config/picochip/divmodsi4.S: Remove.
* config/picochip/lib1funcs.S: Remove.
* config/picochip/longjmp.S: Remove.
* config/picochip/lshrsi3.S: Remove.
* config/picochip/lshrsi3.c: Remove.
* config/picochip/parityhi2.S: Remove.
* config/picochip/popcounthi2.S: Remove.
* config/picochip/setjmp.S: Remove.
* config/picochip/subdi3.S: Remove.
* config/picochip/t-picochip: Remove.
* config/picochip/ucmpsi2.S: Remove.
* config/picochip/udivmodhi4.S: Remove.
* config/picochip/udivmodsi4.S: Remove.

From-SVN: r215039

Daily bump.

From-SVN: r215036

control_flow_insn_p takes an insn

gcc/ChangeLog:
2014-09-08  David Malcolm  <dmalcolm@redhat.com>

* basic-block.h (control_flow_insn_p): Strengthen param from
const_rtx to const rtx_insn *.
* cfgbuild.c (control_flow_insn_p): Likewise.

From-SVN: r215028

gcse.c's modify_mem_list is a list of insns

gcc/ChangeLog:
2014-09-08  David Malcolm  <dmalcolm@redhat.com>

* gcse.c (modify_mem_list): Strengthen this variable from
vec<rtx> * to vec<rtx_insn *> *.
(vec_rtx_heap): Strengthen this typedef from vec<rtx> to
vec<rtx_insn *>.
(load_killed_in_block_p): Strengthen local "list" from vec<rtx> to
vec<rtx_insn *>, and local "setter" from rtx to rtx_insn *.
(record_last_mem_set_info): Strengthen param "insn" from rtx to
rtx_insn *.
(record_last_set_info): Likewise for local "last_set_insn".

From-SVN: r215027

re PR tree-optimization/60196 (Incorrect compilation with -fwrapv and -ftree-vectorize)

PR tree-optimization/60196
PR tree-optimization/63189
* gcc.dg/vect/pr63189.c: New test.
* gcc.dg/vect/pr60196-1.c: New test.
* gcc.dg/vect/pr60196-2.c: New test.

From-SVN: r215025

* doc/invoke.texi (MSP430 Options): Add -minrt.

From-SVN: r215021

rs6000.c (special_handling_values): Add SH_SPLAT.

[gcc]

2014-09-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/rs6000.c (special_handling_values):  Add SH_SPLAT.
(rtx_is_swappable_p): Convert UNSPEC cascading ||s to a switch
statement; allow optimization of UNSPEC_VSPLT_DIRECT with special
handling SH_SPLAT.
(adjust_extract): Fix test for VEC_DUPLICATE case; fix adjustment
of extracted lane.
(adjust_splat): New function.
(handle_special_swappables): Call adjust_splat for SH_SPLAT.
(dump_swap_insn_table): Add case for SH_SPLAT.

[gcc/testsuite]

2014-09-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gcc.target/powerpc/swaps-p8-16.c: New test.

From-SVN: r215019

re PR ipa/63196 (FAIL: g++.dg/torture/pr57140.C   -O3 -fomit-frame-pointer  (internal compiler error))

2014-09-08  Richard Biener  <rguenther@suse.de>

PR ipa/63196
* tree-inline.c (copy_loops): The source loop header should
always be non-NULL.
(tree_function_versioning): If loops need fixup after removing
unreachable blocks fix them.
* omp-low.c (simd_clone_adjust): Do not add incr block to
loop under construction.

From-SVN: r215016

[Obvious] Remove unused aarch64_types_cmtst_qualifiers, was breaking bootstrap.

        * config/aarch64/aarch64-builtins.c
        (aarch64_types_cmtst_qualifiers, TYPES_TST): Remove as unused.

From-SVN: r215015

Remove SF_SIZE etc. target macros.

gcc:
* config/i386/cygming.h (TF_SIZE): Remove.
* config/i386/darwin.h (TF_SIZE): Remove.
* config/i386/dragonfly.h (TF_SIZE): Remove.
* config/i386/freebsd.h (TF_SIZE): Remove.
* config/i386/gnu-user-common.h (TF_SIZE): Remove.
* config/i386/openbsdelf.h (TF_SIZE): Remove.
* config/i386/sol2.h (TF_SIZE): Remove.
* config/ia64/hpux.h (XF_SIZE, TF_SIZE): Remove.
* config/ia64/linux.h (TF_SIZE): Remove.
* doc/tm.texi.in (SF_SIZE, DF_SIZE, XF_SIZE, TF_SIZE): Remove.
* doc/tm.texi: Regenerate.
* system.h (SF_SIZE, DF_SIZE, XF_SIZE, TF_SIZE): Poison.

gcc/c-family:
* c-cppbuiltin.c (c_cpp_builtins): Define macros for mantissa
digits of floating-point modes if -fbuilding-libgcc.

libgcc:
* libgcc2.c (SF_SIZE): Change all uses to __LIBGCC_SF_MANT_DIG__.
(DF_SIZE): Change all uses to __LIBGCC_DF_MANT_DIG__.
(XF_SIZE): Change all uses to __LIBGCC_XF_MANT_DIG__.
(TF_SIZE): Change all uses to __LIBGCC_TF_MANT_DIG__.
* libgcc2.h (SF_SIZE): Change to __LIBGCC_SF_MANT_DIG__.  Give
error if not defined and LIBGCC2_HAS_SF_MODE is defined.
(DF_SIZE): Change to __LIBGCC_DF_MANT_DIG__.  Give error if not
defined and LIBGCC2_HAS_DF_MODE is defined.
(XF_SIZE): Change to __LIBGCC_XF_MANT_DIG__.  Give error if not
defined and LIBGCC2_HAS_XF_MODE is defined.
(TF_SIZE): Change to __LIBGCC_TF_MANT_DIG__.  Give error if not
defined and LIBGCC2_HAS_TF_MODE is defined.

From-SVN: r215014

Remove no-longer-needed fp-bit target macros.

gcc:
* defaults.h (LARGEST_EXPONENT_IS_NORMAL, ROUND_TOWARDS_ZERO):
Remove.
* doc/tm.texi.in (ROUND_TOWARDS_ZERO, LARGEST_EXPONENT_IS_NORMAL):
Remove.
* doc/tm.texi: Regenerate.
* system.h (LARGEST_EXPONENT_IS_NORMAL, ROUND_TOWARDS_ZERO):
Poison.
* config/arm/arm.h (LARGEST_EXPONENT_IS_NORMAL): Remove.
* config/cris/cris.h (__make_dp): Remove.

libgcc:
* fp-bit.c (pack_d, unpack_d): Remove LARGEST_EXPONENT_IS_NORMAL
and ROUND_TOWARDS_ZERO conditionals.

From-SVN: r215013

re PR bootstrap/63204 (gtype-desc.c:887:40: error: 'struct loop' has no member named 'former_header' breaks bootstrap)

2014-09-08  Richard Biener  <rguenther@suse.de>

PR bootstrap/63204
* cfgloop.c (mark_loop_for_removal): Track former header
unconditionally.
* cfgloop.h (struct loop): Add former_header member unconditionally.
* loop-init.c (fix_loop_structure): Enable bogus loop removal
diagnostic unconditionally.

From-SVN: r215012

Daily bump.

From-SVN: r215008

libgcov-interface.c (STRONG_ALIAS): Rename to ...

* libgcov-interface.c (STRONG_ALIAS): Rename to ...
(ALIAS_weak): ... here. Use forwarding function.  Adjust uses.

From-SVN: r215005

Fix PR63190

From-SVN: r215004

re PR rtl-optimization/62208 (ICE with -fwhole-program on valid code at -O3 on x86_64-linux-gnu in trunc_int_for_mode, at explow.c:56)

gcc/
PR rtl-optimization/62208
* simplify-rtx.c (simplify_relational_operation_1): Use CONST0_RTX
rather than const0_rtx in eq/ne-xor simplifications.

gcc/testsuite/
* gcc.target/i386/pr62208.c: New test.

From-SVN: r215002