tree-optimization/97558 - avoid SLP analyzing irrelevant stmts

This avoids analyzing reductions that are not relevant (thus dead)
which eventually will lead into crashes because the participating
stmts meta is not analyzed.  For this to work the patch also
properly removes reduction groups that are not uniformly recognized
as patterns.

2020-11-02  Richard Biener  <rguenther@suse.de>

PR tree-optimization/97558
* tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns):
Check for any mismatch in pattern vs. non-pattern and dissolve
the group if there is one.
* tree-vect-slp.c (vect_analyze_slp_instance): Avoid
analyzing not relevant reductions.
(vect_analyze_slp): Avoid analyzing not relevant reduction
groups.

* gcc.dg/vect/pr97558.c: New testcase.

tree-optimization/97650 - fix ICE in vect_get_and_check_slp_defs

I was mistaken to treat vect_external_def as only applying to
SSA_NAME defs, so check for that.

2020-11-02  Richard Biener  <rguenther@suse.de>

PR tree-optimization/97650
* tree-vect-slp.c (vect_get_and_check_slp_defs): Check
for SSA_NAME before checking SSA_NAME_IS_DEFAULT_DEF.

* gcc.dg/vect/bb-slp-pr97650.c: New testcase.

RISC-V: Check multiletter extension has more than 1 letter

gcc/ChangeLog:

* common/config/riscv/riscv-common.c
(riscv_subset_list::parse_multiletter_ext): Checking multiletter
extension has more than 1 letter.

gcc/testsuite/ChangeLog

* gcc.target/riscv/arch-7.c: New.
* gcc.target/riscv/attribute-10.c: Update test arch string.

RISC-V: Add configure option: --with-multilib-generator to flexible config multi-lib settings.

 - Able to configure complex multi-lib rule in configure time, without modify
   any in-tree source.

 - I was consider to implmenet this into `--with-multilib-list` option,
   but I am not sure who will using that with riscv*-*-elf*, so I decide to
   using another option name for that.

 - --with-multilib-generator will pass arguments to multilib-generator, and
   then using the generated multi-lib config file to build the toolchain.

   e.g. Build riscv gcc, default arch/abi is rv64gc/lp64, and build multilib
       for rv32imafd/ilp32 and rv32i/ilp32; rv32ic/ilp32 will reuse
       rv32i/ilp32.
    $ <GCC-SRC>/configure \
       --target=riscv64-elf \
       --with-arch=rv64gc --with-abi=lp64 \
       --with-multilib-generator=rv32i-ilp32--c;rv32imafd-ilp32--

V3 Changes:

 - Rename --with-multilib-config to --with-multilib-generator
 - Check --with-multilib-generator and --with-multilib-list can't be used at
   same time.

V2 Changes:

 - Fix --with-multilib-config hanling on non riscv*-*-elf* triple.

gcc/ChangeLog:

* config.gcc (riscv*-*-*): Handle --with-multilib-generator.
* configure: Regen.
* configure.ac: Add --with-multilib-generator.
* config/riscv/multilib-generator: Exit when parsing arch string error.
* config/riscv/t-withmultilib-generator: New.
* doc/install.texi: Document --with-multilib-generator.

arm: Improve handling of relocations with small offsets with -mpure-code on v6m (PR96770)

With -mpure-code on v6m (thumb-1), we can use small offsets with
upper/lower relocations to avoid the extra addition of the
offset.

This patch accepts expressions symbol+offset as legitimate constants
when the literal pool is disabled, making sure that the offset is
within the range supported by thumb-1 [0..255] as described in the
AAELF32 documentation.

It also makes sure that thumb1_movsi_insn emits an error in case we
try to use it with an unsupported RTL construct.

2020-09-28  Christophe Lyon  <christophe.lyon@linaro.org>

gcc/
PR target/96770
* config/arm/arm.c (thumb_legitimate_constant_p): Accept
(symbol_ref + addend) when literal pool is disabled.
(arm_valid_symbolic_address_p): Add support for thumb-1 without
MOVT/MOVW.
* config/arm/thumb1.md (*thumb1_movsi_insn): Accept (symbol_ref +
addend) in the pure-code alternative.

gcc/testsuite/
PR target/96770
* gcc.target/arm/pure-code/pr96770.c: New test.

arm: Avoid indirection with -mpure-code on v6m (PR96967)

With -mpure-code on v6m (thumb-1), to avoid a useless indirection when
building the address of a symbol, we want to consider SYMBOL_REF as a
legitimate constant. This way, we build the address using a series of
upper/lower relocations instead of loading the address from memory.

This patch also fixes a missing "clob" conds attribute for
thumb1_movsi_insn, needed because that alternative clobbers the flags.

2020-11-02  Christophe Lyon  <christophe.lyon@linaro.org>

gcc/
PR target/96967
* config/arm/arm.c (thumb_legitimate_constant_p): Add support for
disabled literal pool in thumb-1.
* config/arm/thumb1.md (thumb1_movsi_symbol_ref): Remove.
(*thumb1_movsi_insn): Add support for SYMBOL_REF with -mpure-code.

gcc/testsuite
PR target/96967
* gcc.target/arm/pure-code/pr96767.c: New test.

Darwin: Adjust the PCH area to allow for 16384byte page size.

Newer versions of Darwin report pagesize 20 which means that we
need to adjust the aligment of the PCH area.

gcc/ChangeLog:

* config/host-darwin.c: Align pch_address_space to 16384.

Objective-C : Implement SEL as a built-in typedef.

The reference implementation for Objective-C provides the SEL
typedef (although it is also available from <objc/objc.h>).

gcc/objc/ChangeLog:

* objc-act.c (synth_module_prologue): Get the SEL identifier.
* objc-act.h (enum objc_tree_index): Add OCTI_SEL_NAME.
(objc_selector_name): New.
(SEL_TYPEDEF_NAME): New.
* objc-gnu-runtime-abi-01.c
(gnu_runtime_01_initialize): Initialize SEL typedef.
* objc-next-runtime-abi-01.c
(next_runtime_01_initialize): Likewise.
* objc-next-runtime-abi-02.c

gcc/testsuite/ChangeLog:

* obj-c++.dg/SEL-typedef.mm: New test.
* objc.dg/SEL-typedef.m: New test.

Objective-C/C++ : Improve '@' keyword locations.

When we are lexing tokens for Objective-C, we combine '@' tokens
with a following keyword (when that keyword is a valid Objective-C
one or, for Objective-C, one of the C++ keywords that can appear in
this position).  The responsibility is passed on to the parser to
validate the resulting combination.

The combination of tokens was being done without applying the rule
to their locations - so that we get:

@property
^

instead of what the user might expect:

@property
^~~~~~~~~

This patch combines the source range of the keyword with that of the
'@' sign - which improves diagnostics.

gcc/c-family/ChangeLog:

* c-lex.c (c_lex_with_flags): When combining '@' with a
keyword for Objective-C, combine the location ranges too.

Objective-C++ : Address a FIXME.

We can avoid the spurious additional complaint about a closing
')' by short-circuiting the test in the case we know there's a
syntax error already reported.

gcc/cp/ChangeLog:

* parser.c (cp_parser_objc_at_property_declaration): Use any
exisiting syntax error to suppress complaints about a missing
closing parenthesis in parsing property attributes.

gcc/testsuite/ChangeLog:

* obj-c++.dg/property/at-property-1.mm: Adjust test after
fixing spurious error output.

i386: Set the stack usage to 0 for naked functions

gcc/ChangeLog

* config/i386/i386.c (ix86_expand_prologue): Set the stack usage to 0
for naked functions.

ipa: Fix segmentation fault in function_summary<clone_info*>::get(cgraph_node*)

PR 97660 occurs when cgraph_node::get returns NULL, and this NULL
cgraph_node is then passed to clone_info::get.  As the original assert
prior to the regressing change in r11-4587 allowed for the cgraph_node
to be NULL, clone_info::get is now only called when cgraph_node::get
returns a nonnull value.

gcc/ChangeLog:

PR ipa/97660
* cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Don't call
clone_info::get when cgraph_node::get returns NULL.

testsuite, X86 : Add target requires masm_intel to three tests.

These tests currently fail on targets without Intel assembler support.

gcc/testsuite/ChangeLog:

* gcc.target/i386/amxbf16-asmintel-1.c: Require masm_intel.
* gcc.target/i386/amxint8-asmintel-1.c: Likewise.
* gcc.target/i386/amxtile-asmintel-1.c: Likewise.

libstdc++: Define type traits for wchar_t even when libc support missing

This meets the requirement that std::is_integral_v<wchar_t> is true,
even when full library support for wchar_t via specializations of
char_traits etc. is not provided. This is done by checking
__WCHAR_TYPE__ to see if the compiler knows about the type, rather than
checking the library's own _GLIBCXX_USE_WCHAR_T autoconf macro.

This assumes that the C++ compiler correctly defines wchar_t as a
distinct type, not a typedef for one of the other integeral types. This
is always true for G++ and should be true for any supported non-GNU
compilers.

Similarly, the std::make_unsigned and std::make_signed traits and the
internal helpers std::__is_integer and std::__is_char are also changed
to depend on the same macro.

libstdc++-v3/ChangeLog:

* include/std/type_traits (is_integral<wchar_t>)
(make_unsigned<wchar_t>, make_signed<wchar_t>): Define based
on #ifdef __WCHAR_TYPE__ instead of _GLIBCXX_USE_WCHAR_T.
* include/bits/cpp_type_traits.h (__is_integer<wchar_t>)
(__is_char<wchar_t>): Likewise.

libstdc++: Fix gnu-version-namespace buid

Co-authored-by: Jonathan Wakely <jwakely@redhat.com>
libstdc++-v3/ChangeLog

* src/c++17/floating_from_chars.cc (_GLIBCXX_USE_CX11_ABI): Add define.
(buffering_string): New.
[!_GLIBCXX_USE_CXX11_ABI](reserve_string): New.
(from_chars): Adapt.
* src/c++20/sstream-inst.cc: Limit instantiations to
_GLIBCXX_USE_CXX11_ABI.

libstdc++: Prefer double to long double in std::shuffle_order_engine

The transition algorithm for std::shuffle_order_engine uses long double
to ensure that the value (max() - min() + 1) can be accurately
represented, to avoid bias in the shuffling. However, when the base
engine's range is small enough we can avoid slower long double
arithmetic by using double. For example, long double is unnecessary for
any base engine returning 32-bit values.

This makes std::knuth_b::operator() about 15% faster on x86_64, and
probably even more on targets where long double uses soft-float.

libstdc++-v3/ChangeLog:

* include/bits/random.h (independent_bit_engine): Fix typo
in comment.
(shuffle_order_engine): Fix incorrect description in comment.
* include/bits/random.tcc (__representable_as_double
(__p1_representable_as_double): New helper functions.
(shuffle_order_engine::operator()): Use double for calculation
if (max() - min() + 1) is representable as double.
* testsuite/26_numerics/random/pr60037-neg.cc: Adjust dg-error
line number.

Move clone_info to summary

* Makefile.in: (OBJS): Add symtab-clones.o
(GTFILES): Add symtab-clones.h
* cgraph.c: Include symtab-clones.h.
(cgraph_edge::resolve_speculation): Fix formating
(cgraph_edge::redirect_call_stmt_to_callee): Update.
(cgraph_update_edges_for_call_stmt): Update
(release_function_body): Fix formating.
(cgraph_node::remove): Fix formating.
(cgraph_node::dump): Fix formating.
(cgraph_node::get_availability): Fix formating.
(cgraph_node::call_for_symbol_thunks_and_aliases): Fix formating.
(set_const_flag_1): Fix formating.
(set_pure_flag_1): Fix formating.
(cgraph_node::can_remove_if_no_direct_calls_p): Fix formating.
(collect_callers_of_node_1): Fix formating.
(clone_of_p): Update.
(cgraph_node::verify_node): Update.
(cgraph_c_finalize): Call clone_info::release ().
* cgraph.h (struct cgraph_clone_info): Move to symtab-clones.h.
(cgraph_node): Remove clone_info.
(symbol_table): Add m_clones.
* cgraphclones.c: Include symtab-clone.h.
(duplicate_thunk_for_node): Update.
(cgraph_node::create_clone): Update.
(cgraph_node::create_virtual_clone): Update.
(cgraph_node::find_replacement): Update.
(cgraph_node::materialize_clone): Update.
* gengtype.c (open_base_files): Include symtab-clones.h.
* ipa-cp.c: Include symtab-clones.h.
(initialize_node_lattices): Update.
(want_remove_some_param_p): Update.
(create_specialized_node): Update.
* ipa-fnsummary.c: Include symtab-clones.h.
(ipa_fn_summary_t::duplicate): Update.
* ipa-modref.c: Include symtab-clones.h.
(update_signature): Update.
* ipa-param-manipulation.c: Include symtab-clones.h.
(ipa_param_body_adjustments::common_initialization): Update.
* ipa-prop.c: Include symtab-clones.h.
(adjust_agg_replacement_values): Update.
(ipcp_get_parm_bits): Update.
(ipcp_update_bits): Update.
(ipcp_update_vr): Update.
* ipa-sra.c: Include symtab-clones.h.
(process_isra_node_results): Update.
(disable_unavailable_parameters): Update.
* lto-cgraph.c: Include symtab-clone.h.
(output_cgraph_opt_summary_p): Update.
(output_node_opt_summary): Update.
(input_node_opt_summary): Update.
* symtab-clones.cc: New file.
* symtab-clones.h: New file.
* tree-inline.c (expand_call_inline): Update.
(update_clone_info): Update.
(tree_function_versioning): Update.

Handle fnspec in local ipa-modref

* ipa-modref.c (modref_summary::dump): Dump writes_errno.
(parm_map_for_arg): Break out from ...
(merge_call_side_effects): ... here.
(get_access_for_fnspec): New function.
(process_fnspec): New function.
(analyze_call): Use it.
(analyze_stmt): Update.
(analyze_function): Initialize writes_errno.
(modref_summaries::duplicate): Duplicate writes_errno.
* ipa-modref.h (struct modref_summary): Add writes_errno.
* tree-ssa-alias.c (call_may_clobber_ref_p_1): Check errno.

libstdc++: Use double for unordered container load factors [PR 96958]

My previous commit for this PR changed the types from long double to
double, but didn't change the uses of __builtin_ceill and
__builtin_floorl. It also failed to change the non-inline functions in
src/c++11/hashtable_c++0x.cc. This should fix it properly now.

libstdc++-v3/ChangeLog:

PR libstdc++/96958
* include/bits/hashtable_policy.h (_Prime_rehash_policy)
(_Power2_rehash_policy): Use ceil and floor instead of ceill and
floorl.
* src/c++11/hashtable_c++0x.cc (_Prime_rehash_policy): Likewise.
Use double instead of long double.

libstdc++: Don't initialize from *this inside some views [PR97600]

This works around a subtle issue where instantiating the begin()/end()
member of some views (as part of return type deduction) inadvertently
requires computing the satisfaction value of range<foo_view>.

This is problematic because the constraint range<foo_view> requires the
begin()/end() member to be callable.  But it's not callable until we've
deduced its return type, so evaluation of range<foo_view> yields false
at this point.  And if after both members are instantiated (and their
return types deduced) we evaluate range<foo_view> again, this time it
will yield true since the begin()/end() members are now both callable.
This makes the program ill-formed according to [temp.constr.atomic]/3:

  If, at different points in the program, the satisfaction result is
  different for identical atomic constraints and template arguments, the
  program is ill-formed, no diagnostic required.

The views affected by this issue are those whose begin()/end() member
has a placeholder return type and that member initializes an _Iterator
or _Sentinel object from a reference to *this.  The second condition is
relevant because it means explicit conversion functions are considered
during overload resolution (as per [over.match.copy], I think), and
therefore it causes g++ to check the constraints of the conversion
function view_interface<foo_view>::operator bool().  And this conversion
function's constraints indirectly require range<foo_view>.

This issue is observable on trunk only with basic_istream_view (as in
the testcase in the PR).  But a pending patch that makes g++ memoize
constraint satisfaction values indefinitely (it currently invalidates
the satisfaction cache on various events) causes many existing tests for
the other affected views to fail, because range<foo_view> then remains
false for the whole compilation.

This patch works around this issue by adjusting the constructors of the
_Iterator and _Sentinel types of the affected views to take their
foo_view argument by pointer instead of by reference, so that g++ no
longer considers explicit conversion functions when resolving the
direct-initialization inside these views' begin()/end() members.

libstdc++-v3/ChangeLog:

PR libstdc++/97600
* include/std/ranges (basic_istream_view::begin): Initialize
_Iterator from 'this' instead of '*this'.
(basic_istream_view::_Iterator::_Iterator): Adjust constructor
accordingly.
(filter_view::_Iterator::_Iterator): Take a filter_view*
argument instead of a filter_view& argument.
(filter_view::_Sentinel::_Sentinel): Likewise.
(filter_view::begin): Initialize _Iterator from 'this' instead
of '*this'.
(filter_view::end): Likewise.
(transform_view::_Iterator::_Iterator): Take a _Parent* instead
of a _Parent&.
(filter_view::_Iterator::operator+): Adjust accordingly.
(filter_view::_Iterator::operator-): Likewise.
(filter_view::begin): Initialize _Iterator from 'this' instead
of '*this'.
(filter_view::end): Likewise.
(join_view::_Iterator): Take a _Parent* instead of a _Parent&.
(join_view::_Sentinel): Likewise.
(join_view::begin): Initialize _Iterator from 'this' instead of
'*this'.
(join_view::end): Initialize _Sentinel from 'this' instead of
'*this'.
(split_view::_OuterIter): Take a _Parent& instead of a _Parent*.
(split_view::begin): Initialize _OuterIter from 'this' instead
of '*this'.
(split_view::end): Likewise.
* testsuite/std/ranges/97600.cc: New test.

libstdc++: Implement P2017R1 "Conditionally borrowed ranges"

This makes some range adaptors model the borrowed_range concept if they
are adapting a borrowed range. This hasn't been added to the C++23
working paper yet, but it has been approved by LWG, and the
recommendation is to treat it as a defect report for C++20 as well.

libstdc++-v3/ChangeLog:

* include/std/ranges (enable_borrowed_view<take_view<T>>)
(enable_borrowed_view<drop_view<T>>)
(enable_borrowed_view<drop_while_view<T>>)
(enable_borrowed_view<reverse_view<T>>)
(enable_borrowed_view<common_view<T>>)
(enable_borrowed_view<elements_view<T>>): Add partial
specializations as per P2017R1.
* testsuite/std/ranges/adaptors/conditionally_borrowed.cc:
New test.

PowerPC: Don't assume all targets have GLIBC.

gcc/
2020-10-30  Michael Meissner  <meissner@linux.ibm.com>

* config/rs6000/rs6000.c (glibc_supports_ieee_128bit): New helper
function.
(rs6000_option_override_internal): Call it.

libstdc++: Use double for unordered container load factors [PR 96958]

These calculations were changed to use long double nearly ten years ago
in order to get more precision than float:
https://gcc.gnu.org/pipermail/libstdc++/2011-September/036420.html

However, double should be sufficient, whlie being potentially faster
than long double, and not requiring soft FP calculations for targets
without native long double support.

libstdc++-v3/ChangeLog:

PR libstdc++/96958
* include/bits/hashtable_policy.h (_Prime_rehash_policy)
(_Power2_rehash_policy): Use double instead of long double.

libstdc++: Fix some more warnings in test

libstdc++-v3/ChangeLog:

* testsuite/23_containers/vector/bool/modifiers/insert/31370.cc:
Avoid -Wcatch-value warnings.

PR libfortran/97581 - clean up size calculation of random generator state

The random number generator internal state may be saved to/restored from
an array of integers.  Clean up calculation of needed number of elements
to avoid redefiniton of auxiliary macro SZ.

libgfortran/ChangeLog:

* intrinsics/random.c (SZ_IN_INT_4): Define size of state in int32_t.
(SZ_IN_INT_8): Define size of state in int64_t.
(SZ): Remove.
(random_seed_i4): Use size SZ_IN_INT_4 instead of SZ.
(random_seed_i8): Use size SZ_IN_INT_8 instead of SZ.

Add -fzero-call-used-regs option and zero_call_used_regs function attributes.

This new feature causes the compiler to zero a  subset of all call-used
registers at function return.  This is used to increase program security
by either mitigating Return-Oriented Programming (ROP) attacks or
preventing information leakage through registers.

gcc/ChangeLog:

2020-10-30  Qing Zhao  <qing.zhao@oracle.com>
    H.J.Lu  <hjl.tools@gmail.com>

* common.opt: Add new option -fzero-call-used-regs
* config/i386/i386.c (zero_call_used_regno_p): New function.
(zero_call_used_regno_mode): Likewise.
(zero_all_vector_registers): Likewise.
(zero_all_st_registers): Likewise.
(zero_all_mm_registers): Likewise.
(ix86_zero_call_used_regs): Likewise.
(TARGET_ZERO_CALL_USED_REGS): Define.
* df-scan.c (df_epilogue_uses_p): New function.
(df_get_exit_block_use_set): Replace EPILOGUE_USES with
df_epilogue_uses_p.
* df.h (df_epilogue_uses_p): Declare.
* doc/extend.texi: Document the new zero_call_used_regs attribute.
* doc/invoke.texi: Document the new -fzero-call-used-regs option.
* doc/tm.texi: Regenerate.
* doc/tm.texi.in (TARGET_ZERO_CALL_USED_REGS): New hook.
* emit-rtl.h (struct rtl_data): New field must_be_zero_on_return.
* flag-types.h (namespace zero_regs_flags): New namespace.
* function.c (gen_call_used_regs_seq): New function.
(class pass_zero_call_used_regs): New class.
(pass_zero_call_used_regs::execute): New function.
(make_pass_zero_call_used_regs): New function.
* optabs.c (expand_asm_reg_clobber_mem_blockage): New function.
* optabs.h (expand_asm_reg_clobber_mem_blockage): Declare.
* opts.c (zero_call_used_regs_opts): New structure array
initialization.
(parse_zero_call_used_regs_options): New function.
(common_handle_option): Handle -fzero-call-used-regs.
* opts.h (zero_call_used_regs_opts): New structure array.
* passes.def: Add new pass pass_zero_call_used_regs.
* recog.c (valid_insn_p): New function.
* recog.h (valid_insn_p): Declare.
* resource.c (init_resource_info): Replace EPILOGUE_USES with
df_epilogue_uses_p.
* target.def (zero_call_used_regs): New hook.
* targhooks.c (default_zero_call_used_regs): New function.
* targhooks.h (default_zero_call_used_regs): Declare.
* tree-pass.h (make_pass_zero_call_used_regs): Declare.

gcc/c-family/ChangeLog:

2020-10-30  Qing Zhao  <qing.zhao@oracle.com>
    H.J.Lu  <hjl.tools@gmail.com>

* c-attribs.c (c_common_attribute_table): Add new attribute
zero_call_used_regs.
(handle_zero_call_used_regs_attribute): New function.

gcc/testsuite/ChangeLog:

2020-10-30  Qing Zhao  <qing.zhao@oracle.com>
    H.J.Lu  <hjl.tools@gmail.com>

* c-c++-common/zero-scratch-regs-1.c: New test.
* c-c++-common/zero-scratch-regs-10.c: New test.
* c-c++-common/zero-scratch-regs-11.c: New test.
* c-c++-common/zero-scratch-regs-2.c: New test.
* c-c++-common/zero-scratch-regs-3.c: New test.
* c-c++-common/zero-scratch-regs-4.c: New test.
* c-c++-common/zero-scratch-regs-5.c: New test.
* c-c++-common/zero-scratch-regs-6.c: New test.
* c-c++-common/zero-scratch-regs-7.c: New test.
* c-c++-common/zero-scratch-regs-8.c: New test.
* c-c++-common/zero-scratch-regs-9.c: New test.
* c-c++-common/zero-scratch-regs-attr-usages.c: New test.
* gcc.target/i386/zero-scratch-regs-1.c: New test.
* gcc.target/i386/zero-scratch-regs-10.c: New test.
* gcc.target/i386/zero-scratch-regs-11.c: New test.
* gcc.target/i386/zero-scratch-regs-12.c: New test.
* gcc.target/i386/zero-scratch-regs-13.c: New test.
* gcc.target/i386/zero-scratch-regs-14.c: New test.
* gcc.target/i386/zero-scratch-regs-15.c: New test.
* gcc.target/i386/zero-scratch-regs-16.c: New test.
* gcc.target/i386/zero-scratch-regs-17.c: New test.
* gcc.target/i386/zero-scratch-regs-18.c: New test.
* gcc.target/i386/zero-scratch-regs-19.c: New test.
* gcc.target/i386/zero-scratch-regs-2.c: New test.
* gcc.target/i386/zero-scratch-regs-20.c: New test.
* gcc.target/i386/zero-scratch-regs-21.c: New test.
* gcc.target/i386/zero-scratch-regs-22.c: New test.
* gcc.target/i386/zero-scratch-regs-23.c: New test.
* gcc.target/i386/zero-scratch-regs-24.c: New test.
* gcc.target/i386/zero-scratch-regs-25.c: New test.
* gcc.target/i386/zero-scratch-regs-26.c: New test.
* gcc.target/i386/zero-scratch-regs-27.c: New test.
* gcc.target/i386/zero-scratch-regs-28.c: New test.
* gcc.target/i386/zero-scratch-regs-29.c: New test.
* gcc.target/i386/zero-scratch-regs-30.c: New test.
* gcc.target/i386/zero-scratch-regs-31.c: New test.
* gcc.target/i386/zero-scratch-regs-3.c: New test.
* gcc.target/i386/zero-scratch-regs-4.c: New test.
* gcc.target/i386/zero-scratch-regs-5.c: New test.
* gcc.target/i386/zero-scratch-regs-6.c: New test.
* gcc.target/i386/zero-scratch-regs-7.c: New test.
* gcc.target/i386/zero-scratch-regs-8.c: New test.
* gcc.target/i386/zero-scratch-regs-9.c: New test.

Take insn scratch RA requirements into account in IRA.

  The patch changes insn scratches which require registers for all
insn alternatives (in other words w/o X constraint in scratch
constraint string).  This is done before IRA staring its work.  LRA
still continue to change the rest scratches (with X constraint and in
insn created during IRA) into pseudos.  As before the patch at the end
of LRA work, spilled scratch pseudos (for which X constraint was
chosen) changed into scratches back.

gcc/ChangeLog:

* lra.c (get_scratch_reg): New function.
(remove_scratches_1): Rename remove_insn_scratches.  Use
ira_remove_insn_scratches and get_scratch_reg.
(remove_scratches): Do not
initialize scratches, scratch_bitmap, and scratch_operand_bitmap.
(lra): Call ira_restore_scratches instead of restore_scratches.
(struct sloc, sloc_t, scratches, scratch_bitmap)
(scratch_operand_bitmap, lra_former_scratch_p)
(lra_former_scratch_operand_p, lra_register_new_scratch_op)
(restore_scratches): Move them to ...
* ira.c: ... here.
(former_scratch_p, former_scratch_operand_p): Rename to
ira_former_scratch_p and ira_former_scratch_operand_p.
(contains_X_constraint_p): New function.
(register_new_scratch_op): Rename to ira_register_new_scratch_op.
Change it to work for IRA and LRA.
(restore_scratches): Rename to ira_restore_scratches.
(get_scratch_reg, ira_remove_insn_scratches): New functions.
(ira): Call ira_remove_scratches if we use LRA.
* ira.h (ira_former_scratch_p, ira_former_scratch_operand_p): New
prototypes.
(ira_register_new_scratch_op, ira_restore_scratches): New prototypes.
(ira_remove_insn_scratches): New prototype.
* lra-int.h (lra_former_scratch_p, lra_former_scratch_operand_p):
Remove prototypes.
(lra_register_new_scratch_op): Ditto.
* lra-constraints.c: Rename lra_former_scratch_p and
lra_former_scratch_p to ira_former_scratch_p and to
ira_former_scratch_p.
* lra-remat.c: Ditto.
* lra-spills.c: Rename lra_former_scratch_p to ira_former_scratch_p.

PR middle-end/97556 - ICE on excessively large index into a multidimensional array

gcc/ChangeLog:

PR middle-end/97556
* builtins.c (access_ref::add_offset): Cap offset lower bound
to at most the the upper bound.

gcc/testsuite/ChangeLog:

PR middle-end/97556
* gcc.dg/Warray-bounds-70.c: New test.

libstdc++: Fix the default constructor of ranges::__detail::__box

The class template semiregular-box<T> of [range.semi.wrap] is specified
to value-initialize the underlying object whenever its type is default
initializable.  Our primary template for __detail::__box respects this
requirement, but the recently added partial specialization (for types
that are already semiregular) does not.

This patch fixes this issue, and additionally makes the corresponding in
place constructor explicit (as in the primary template).

libstdc++-v3/ChangeLog:

* include/std/ranges (__detail::__box): For the partial
specialization used by types that are already semiregular,
make the default constructor value-initialize the underlying
object instead of default-initializing it.  Make its in place
constructor explicit.
* testsuite/std/ranges/adaptors/detail/semiregular_box.cc:
Augment test.

testsuite: Avoid TCL errors when rootme or ASAN/TSAN/UBSAN is not avail

gcc/testsuite/
* g++.dg/guality/guality.exp: Skip $rootme-based check if unset.
* gcc.dg/guality/guality.exp: Likewise.
* gfortran.dg/guality/guality.exp: Likewise.
* lib/asan-dg.exp: Don't use $asan_saved_library_path if not set.
* lib/tsan-dg.exp: Don't use $tsan_saved_library_path if not set.
* lib/ubsan-dg.exp: Don't use $ubsan_saved_library_path if not set.

Fortran: Update omp atomic for OpenMP 5

gcc/fortran/ChangeLog:

* dump-parse-tree.c (show_omp_clauses): Handle atomic clauses.
(show_omp_node): Call it for atomic.
* gfortran.h (enum gfc_omp_atomic_op): Add GFC_OMP_ATOMIC_UNSET,
remove GFC_OMP_ATOMIC_SEQ_CST and GFC_OMP_ATOMIC_ACQ_REL.
(enum gfc_omp_memorder): Replace OMP_MEMORDER_LAST by
OMP_MEMORDER_UNSET, add OMP_MEMORDER_SEQ_CST/OMP_MEMORDER_RELAXED.
(gfc_omp_clauses): Add capture and atomic_op.
(gfc_code): remove omp_atomic.
* openmp.c (enum omp_mask1): Add atomic, capture, memorder clauses.
(gfc_match_omp_clauses): Match them.
(OMP_ATOMIC_CLAUSES): Add.
(gfc_match_omp_flush): Update for 'last' to 'unset' change.
(gfc_match_omp_oacc_atomic): Removed and placed content ..
(gfc_match_omp_atomic): ... here. Update for OpenMP 5 clauses.
(gfc_match_oacc_atomic): Match directly here.
(resolve_omp_atomic, gfc_resolve_omp_directive): Update.
* parse.c (parse_omp_oacc_atomic): Update for struct gfc_code changes.
* resolve.c (gfc_resolve_blocks): Update assert.
* st.c (gfc_free_statement): Also call for EXEC_O{ACC,MP}_ATOMIC.
* trans-openmp.c (gfc_trans_omp_atomic): Update.
(gfc_trans_omp_flush): Update for 'last' to 'unset' change.

gcc/testsuite/ChangeLog:

* gfortran.dg/gomp/atomic-2.f90: New test.
* gfortran.dg/gomp/atomic.f90: New test.

Fix thunk info WRT PCH

PR pch/97593
* cgraph.c (cgraph_node::create_thunk): Register thunk as early during
parsing.
* cgraphunit.c (analyze_functions): Call
thunk_info::process_early_thunks.
* symtab-thunks.cc (struct unprocessed_thunk): New struct.
(thunks): New static variable.
(thunk_info::register_early): New member function.
(thunk_info::process_early_thunks): New member function.
* symtab-thunks.h (thunk_info::register_early): Declare.
(thunk_info::process_early_thunks): Declare.

Disable TBAA for array descriptors.

* trans-types.c: Include alias.h
(gfc_get_array_type_bounds): Set typeless storage.

tree-optimization/97623 - avoid excessive insert iteration for hoisting

This avoids requiring insert iteration for back-to-back hoisting
opportunities as seen in the added testcase.  For the PR at hand
this halves the number of insert iterations retaining only
the hard to avoid PRE / hoist insert back-to-backs.

2020-10-30  Richard Biener  <rguenther@suse.de>

PR tree-optimization/97623
* tree-ssa-pre.c (insert): First do hoist insertion in
a backward walk.

* gcc.dg/tree-ssa/ssa-hoist-7.c: New testcase.

tree-optimization/97626 - handle SCCs properly in SLP stmt analysis

This makes sure to roll-back the whole SCC when we fail stmt
analysis, otherwise the optimistic visited treatment breaks down
with different entries.  Rollback is easy when tracking additions
to visited in a vector which also makes the whole thing cheaper
than the two hash-sets used before.

2020-10-30  Richard Biener  <rguenther@suse.de>

PR tree-optimization/97626
* tree-vect-slp.c (vect_slp_analyze_node_operations):
Exchange the lvisited hash-set for a vector, roll back
recursive adds to visited when analysis failed.
(vect_slp_analyze_operations): Likewise.

* gcc.dg/vect/bb-slp-pr97626.c: New testcase.

openmp: Use FIELD_TGT_EMPTY once more

2020-10-30  Jakub Jelinek  <jakub@redhat.com>

* target.c (gomp_map_vars_internal): Use FIELD_TGT_EMPTY macro
even in field_tgt_clear initializer.

[OpenACC] Elaborate testcases that verify column location information [PR92793]

After PR92793 commit 9c81750c5bedd7883182ee2684a012c6210ebe1d "Fortran] PR
92793 - fix column used for error diagnostic", commit
d0d0ba20f2345023e9cec2419c9fb9e6cc7098c6 did "Add tests to verify OpenACC
clause locations", later fixed up in PR92901 commit
e6c90dba73291435c244decb9a89c47019cc5a45 to "Fix PR92901: Change test
expectation for C++ in OpenACC test clause-locations.c".

Now, add some more testing to verify/document the status quo.

gcc/testsuite/
PR fortran/92793
* c-c++-common/goacc/clause-locations.c: Rewrite into...
* c-c++-common/goacc/pr92793-1.c: ... this.
* gfortran.dg/goacc/clause-locations.f90: Rewrite into...
* gfortran.dg/goacc/pr92793-1.f90: ... this.

gcc-changelog: Handle situations like '* tree-vect-slp.c (): '

contrib/ChangeLog:

* gcc-changelog/git_commit.py: Handle empty groups in
file description.
* gcc-changelog/test_email.py: New test.
* gcc-changelog/test_patches.txt: Likewise.

AArch64: Add FLAG for float conversion intrinsics [PR94442]

2020-10-30  Zhiheng Xie  <xiezhiheng@huawei.com>
    Nannan Zheng  <zhengnannan@huawei.com>

gcc/ChangeLog:

* config/aarch64/aarch64-simd-builtins.def: Add proper FLAG
for conversion intrinsics.

tree-optimization/97633 - fix SLP scheduling of single-node cycles

This makes sure to update backedges in single-node cycles.

2020-10-30  Richard Biener  <rguenther@suse.de>

PR tree-optimization/97633
* tree-vect-slp.c (): Update backedges in single-node cycles.
Optimize processing of externals.

* g++.dg/vect/slp-pr97636.cc: New testcase.
* gcc.dg/vect/bb-slp-pr97633.c: Likewise.

aarch64: Fix PR96998 and restore code quality in combine

This change fixes a bug in the AArch64 backend. Currently, we accept an
odd sign_extract representation of addresses, but don't accept that same
odd form of address as an LEA.

This is the cause of PR96998. In the testcase given in the PR, combine
produces:

(insn 9 8 10 3 (set (mem:SI (plus:DI (sign_extract:DI (mult:DI (subreg:DI (reg/v:SI 92 [ g ]) 0)
                        (const_int 4 [0x4]))
                    (const_int 34 [0x22])
                    (const_int 0 [0]))
                (reg/f:DI 96)) [3 *i_5+0 S4 A32])
        (asm_operands:SI ("") ("=Q") 0 []
             []
             [] test.c:11)) "test.c":11:5 -1
     (expr_list:REG_DEAD (reg/v:SI 92 [ g ])
        (nil)))

Then LRA reloads the address and we ICE because we fail to recognize the
sign_extract outside the mem:

(insn 33 8 34 3 (set (reg:DI 100)
        (sign_extract:DI (ashift:DI (subreg:DI (reg/v:SI 92 [ g ]) 0)
                (const_int 2 [0x2]))
            (const_int 34 [0x22])
            (const_int 0 [0]))) "test.c":11:5 -1
     (nil))

The aarch64 changes here remove the support for this sign_extract
representation of addresses, fixing PR96998. Now this by itself would
regress code quality, so this change is paired with an improvement to
combine which prevents an extract rtx from being emitted in this case:
we now write the rtx above as a shift of an extend, which allows the
combination to go ahead.

Prior to this, combine.c:make_extraction() identified where we can emit
an ashift of an extend in place of an extraction, but failed to make the
corresponding canonicalization/simplification when presented with a mult
by a power of two. Such a representation is canonical when representing
a left-shifted address inside a mem.

This change remedies this situation. For rtxes such as:

(mult:DI (subreg:DI (reg:SI r) 0) (const_int 2^n))

where the bottom 32 + n bits are valid (the higher-order bits are
undefined) and make_extraction() is being asked to sign_extract the
lower (valid) bits, after the patch, we rewrite this as:

(mult:DI (sign_extend:DI (reg:SI r)) (const_int 2^n))

instead of using a sign_extract.

gcc/ChangeLog:

PR target/96998
* combine.c (make_extraction): Also handle shifts written as
(mult x 2^n), avoid creating an extract rtx for these.
* config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Delete.
(aarch64_classify_index): Remove extract-based address handling.
(aarch64_strip_extend): Likewise.
(aarch64_rtx_arith_op_extract_p): Likewise, remove now-unused parameter.
Update callers...
(aarch64_rtx_costs): ... here.

gcc/testsuite/ChangeLog:

PR target/96998
* gcc.c-torture/compile/pr96998.c: New test.

Predefine __ppc and __ppc__ for VxWorks 7

Unfortunately, some VxWorks 7r2 system headers rely on a
couple more variations of the predefined macros expected
to characterize a "powerpc" target that we discussed recently.

setjmp.h, for example, relies on __ppc and the absence of
a definition results in "gcc" dejagnu test failures from all
the tests #including that header, which stumble on:

  #error "_JBLEN not set!"

The other case is __ppc__ expected by yvals.h, key to libstdc++.

This change adjusts the VxWorks 7 section of our configuration
to honor those expectations.

2020-10-29  Olivier Hainque  <hainque@adacore.com>

gcc/
* config/rs6000/vxworks.h (TARGET_OS_CPP_BUILTINS): Also
builtin_define __ppc and __ppc__ for VxWorks 7.

Introduce support for vxworks7r2 on x86 and x86_64

This change extends the VxWorks support on intel CPUs to
VxWorks7r2 for x86_64 as well as x86, with a "mcmodel=large"
additional multilib for the 64bit configuration.

The support for fPIC is not functional yet for this model,
so we just don't add the corresponding multilib.

We extend the range of CPU families handled by TARGET_OS_CPP_BUILTINS,
accounting for the fact that archs older than PENTIUM4 are
not supported (any more) by VxWorks 7.

As we did for powerpc, we leverage VX_CPU_PREFIX to emit different
forms of definitions for different families of VxWorks as the system
headers's expectations has evolved between Vx 5, 6 and 7.

2020-10-27  Olivier Hainque  <hainque@adacore.com>

gcc/
* config.gcc: Adjust the ix86/x86_64-wrs-vxworks filters
to apply to VxWorks 7 as well.
* config/i386/t-vxworks (MULTILIB_OPTIONS, MULTILIB_DIRNAMES):
Remove the fPIC multilib and add one for the large code model
on x86_64.
* config/i386/vxworks.h: Separate sections for TARGET_VXWORKS7,
other variants and common bits.
(TARGET_OS_CPP_BUILTINS): Augment to support a range of CPU
families. Leverage VX_CPU_PREFIX.
(CC1_SPEC): Add definition.
(STACK_CHECK_PROTECT): Use conditional expression instead of
heavier to read conditioned macro definitions.

libgcc/
* config.host: Adjust the ix86/x86_64-wrs-vxworks filters
to apply to VxWorks 7 as well.

Co-authored-by: Douglas Rupp <rupp@adacore.com>
Co-authored-by: Pat Bernardi <bernardi@adacore.com>

openmp: Fix handling of allocate clause on taskloop

This patch fixes gimplification of allocate clause on taskloop - puts
allocate on inner taskloop only if there is allocate clause, because otherwise
the data sharing clauses are only on the task construct in the construct sandwich.

2020-10-30  Jakub Jelinek  <jakub@redhat.com>

* gimplify.c (gimplify_scan_omp_clauses): Force
OMP_CLAUSE_ALLOCATE_ALLOCATOR into a temporary if it is non-NULL and
non-constant.
(gimplify_omp_for): Only put allocate on inner taskloop if lastprivate
for the same variable is going to be put there, and in that case
if the OMP_CLAUSE_ALLOCATE_ALLOCATOR is non-NULL non-constant, make
the allocator firstprivate on task.

* c-c++-common/gomp/allocate-3.c: New test.

openmp: Handle non-static data members in allocate clause and other C++ allocate fixes

This allows specification of non-static data members in allocate clause like it
can be specified in other privatization clauses and adds a new testcase that covers
also handling of that clause in templates.

2020-10-30  Jakub Jelinek  <jakub@redhat.com>

* semantics.c (finish_omp_clauses) <case OMP_CLAUSE_ALLOCATE>: Handle
non-static members in methods.
* pt.c (tsubst_omp_clauses): Handle OMP_CLAUSE_ALLOCATE.

* c-c++-common/gomp/allocate-1.c (qux): Add another test.
* g++.dg/gomp/allocate-1.C: New test.

Adjust gcc.dg/vect/bb-slp-pr65935.c

This adjusts the testcase to allow splitting up the group for
larger vector sizes and thus printing the splat message multiple times.

2020-10-30  Richard Biener  <rguenther@suse.de>

* gcc.dg/vect/bb-slp-pr65935.c: Adjust.

PowerPC: Adjust float128/ibm128 warnings.

This patch ccombines two patches:

1) If GLIBC is 2.32 or newer, and the language is C or C++, allow the user to
   change the long double type without having to use -Wno-psabi.

2) Adjust the warnings for intermixing __float128 and __ibm128 to accomidate
   the future change to allow long double to use the IEEE 128-bit format.

gcc/
2020-10-29  Michael Meissner  <meissner@linux.ibm.com>

* config/rs6000/rs6000.c (rs6000_option_override_internal): Allow
long double type to be changed for C/C++ if glibc 2.32 or newer.
(rs6000_invalid_binary_op): Update error messages about mixing IBM
long double and IEEE 128-bit.

gcc/testsuite/
2020-10-27  Michael Meissner  <meissner@linux.ibm.com>

* gcc.target/powerpc/float128-mix-2.c: New test.
* gcc.target/powerpc/float128-mix-3.c: New test.
* gcc.target/powerpc/float128-mix.c: Update failure messages.

libstdc++: AIX xfail for_overwrite.cc testcase

The 20_util/unique_ptr/creation/for_overwrite.cc testcase relies on
operator new, which requires special features on AIX.  This patch
disables the testcase.

libstdc++-v3/ChangeLog:

* testsuite/20_util/unique_ptr/creation/for_overwrite.cc: XFAIL on AIX.

Daily bump.

libstdc++: Fix linker script to remove conflicting patterns

This should fix a bootstrap error on Solaris, due to some of the new
symbols matching old patterns as well as new ones.

libstdc++-v3/ChangeLog:

* config/abi/pre/gnu.ver (GLIBCXX_3.4.21): Tighten patterns
for old <sstream> symbols some more.

libstdc++: Avoid narrowing conversion in subrange constructor

libstdc++-v3/ChangeLog:

* include/bits/ranges_util.h (subrange::subrange(R&&)): Use
direct-initialization instead of list-initialization, so a
potential narrowing conversion from ranges::size(r) to the
stored size isn't ill-formed.

libstdc++: Fix some warnings in headers

These are usually suppressed in system headers, but should be fixed
anyway.

libstdc++-v3/ChangeLog:

* include/bits/parse_numbers.h (_Select_int_base): Avoid
narrowing conversion in constant expression.
* include/experimental/buffer (buffer_copy): Avoid narrowing
conversion.
* include/experimental/internet (hash<>::operator()): Do not
use deprecated 'argument_type' member.
* include/std/variant (variant::emplace): Use cast instead
of implicit conversion from size_t to narrower unsigned type.

libstdc++: Prevent deprecation warnings from <tr1/shared_ptr>

libstdc++-v3/ChangeLog:

* include/tr1/shared_ptr.h (__shared_count, __shared_ptr)
(shared_ptr): Add diagnostic pragmas around uses of auto_ptr.
* testsuite/tr1/2_general_utilities/shared_ptr/cons/43820_neg.cc:
Adust dg-error line numbers.

libstdc++: Avoid warnings in tests

This fixes some warnings emitted when testing with warning flags added.
Some of these are only necessary when testing with -Wsystem-headers, but
either way it cleans up the tests to be less noisy under non-default
flags.

libstdc++-v3/ChangeLog:

* testsuite/18_support/96817.cc: Avoid -Wunused warnings.
* testsuite/20_util/any/assign/2.cc: Likewise.
* testsuite/20_util/any/cons/2.cc: Likewise.
* testsuite/20_util/align/1.cc: Avoid -Wsign-compare warning.
* testsuite/20_util/function/65760.cc: Avoid -Wunused warning.
* testsuite/20_util/function/1.cc: Avoid -Wcatch-value warning.
* testsuite/20_util/function/cons/move_target.cc: Avoid -Wunused
warning.
* testsuite/20_util/headers/memory/synopsis.cc: Add exception
specification.
* testsuite/20_util/monotonic_buffer_resource/allocate.cc: Avoid
-Wsign-compare warning.
* testsuite/20_util/tuple/cons/deduction.cc: Avoid -Wunused
warning.
* testsuite/20_util/specialized_algorithms/uninitialized_copy/808590-cxx11.cc:
Avoid -Wdeprecated-copy warning.
* testsuite/21_strings/basic_string/56166.cc: Avoid
-Wcatch-value warning.
* testsuite/21_strings/basic_string/numeric_conversions/char/stod.cc:
Avoid -Wcatch-value warnings.
* testsuite/21_strings/basic_string/numeric_conversions/char/stof.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stoi.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stol.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stold.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stoll.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stoul.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/char/stoull.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/wchar_t/stod.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/wchar_t/stof.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/wchar_t/stoi.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/wchar_t/stol.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/wchar_t/stold.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/wchar_t/stoll.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/wchar_t/stoul.cc:
Likewise.
* testsuite/21_strings/basic_string/numeric_conversions/wchar_t/stoull.cc:
Likewise.
* testsuite/21_strings/basic_string_view/operations/compare/char/nonnull.cc:
Prune additional diagnostics.
* testsuite/21_strings/basic_string_view/operations/find/char/nonnull.cc:
Likewise.
* testsuite/21_strings/basic_string_view/operations/rfind/char/nonnull.cc:
Likewise.
* testsuite/21_strings/headers/string/synopsis.cc: Add exception
specifications.
* testsuite/22_locale/locale/cons/12352.cc: Define sized
delete operators to avoid warnings.
* testsuite/23_containers/deque/modifiers/swap/1.cc: Add
exception specification.
* testsuite/23_containers/forward_list/cons/11.cc: Avoid
-Wdeprecated-copy warning.
* testsuite/23_containers/headers/bitset/synopsis.cc: Add
exception specification.
* testsuite/23_containers/headers/deque/synopsis.cc: Likewise.
* testsuite/23_containers/headers/forward_list/synopsis.cc:
Likewise.
* testsuite/23_containers/headers/list/synopsis.cc: Likewise.
* testsuite/23_containers/headers/map/synopsis.cc: Likewise.
* testsuite/23_containers/headers/queue/synopsis.cc: Likewise.
* testsuite/23_containers/headers/set/synopsis.cc: Likewise.
* testsuite/23_containers/headers/vector/synopsis.cc: Likewise.
* testsuite/23_containers/list/modifiers/swap/1.cc: Likewise.
* testsuite/23_containers/map/modifiers/swap/1.cc: Likewise.
* testsuite/23_containers/multimap/modifiers/swap/1.cc:
Likewise.
* testsuite/23_containers/multiset/modifiers/swap/1.cc:
Likewise.
* testsuite/23_containers/set/modifiers/swap/1.cc: Likewise.
* testsuite/23_containers/unordered_set/56267-2.cc: Avoid
-Wdeprecated-copy warning.
* testsuite/23_containers/vector/bool/23632.cc: Avoid
-Wempty-body warning.
* testsuite/23_containers/vector/modifiers/swap/1.cc: Add
exception specification.
* testsuite/25_algorithms/heap/moveable2.cc: Fix misplaced
parentheses around arguments.
* testsuite/25_algorithms/sample/1.cc: Use return value.
* testsuite/25_algorithms/search/searcher.cc: Avoid -Wunused
warnings.
* testsuite/27_io/basic_ostream/exceptions/char/9561.cc:
Likewise.
* testsuite/27_io/basic_ostream/exceptions/wchar_t/9561.cc:
Likewise.
* testsuite/27_io/filesystem/operations/remove_all.cc: Avoid
-Wsign-compare warning.
* testsuite/experimental/any/assign/2.cc: Avoid -Wunused warnings.
* testsuite/experimental/any/cons/2.cc: Likewise.
* testsuite/experimental/filesystem/operations/remove_all.cc:
Avoid -Wign-compare warning.
* testsuite/experimental/memory/observer_ptr/cons/cons.cc:
Likewise.
* testsuite/experimental/memory_resource/null_memory_resource.cc:
Likewise.
* testsuite/experimental/source_location/1.cc: Avoid -Waddress
warning.
* testsuite/ext/pod_char_traits.cc: Avoid -Wunused warning.
* testsuite/ext/vstring/modifiers/clear/56166.cc: Avoid
-Wcatch-value.
* testsuite/std/concepts/concepts.lang/concept.swappable/swap.cc:
Avoid -Wunused warning.
* testsuite/std/concepts/concepts.lang/concept.swappable/swappable.cc:
Likewise.
* testsuite/tr1/2_general_utilities/shared_ptr/cons/43820_neg.cc:
Prune additional warnings.
* testsuite/tr1/3_function_objects/function/1.cc: Avoid
-Wcatch-value warning.
* testsuite/util/replacement_memory_operators.h: Define sized
delete to avoid warnings.
* testsuite/util/testsuite_api.h (_NonDefaultConstructible): Add
user-declared assignment operator to stop -Wdeprecated-copy
warnings.
* testsuite/util/testsuite_containers.h: Avoid -Wunused warning.
* testsuite/util/testsuite_iterators.h: Avoid -Wsign-compare
warnings.
* testsuite/util/testsuite_new_operators.h: Define sized deleted.

c++: Implement CWG 625: Use of auto as template-arg [PR97479]

This patch implements CWG 625 which prohibits using auto in a template
argument.  A few tests used this construction.  Since this usage was
allowed by the Concepts TS, we only give an error in C++20.

gcc/cp/ChangeLog:

DR 625
PR c++/97479
* parser.c (cp_parser_type_id_1): Reject using auto as
a template-argument in C++20.

gcc/testsuite/ChangeLog:

DR 625
PR c++/97479
* g++.dg/cpp0x/auto3.C: Update dg-error.
* g++.dg/cpp0x/auto9.C: Likewise.
* g++.dg/cpp2a/concepts-pr84979-2.C: Likewise.
* g++.dg/cpp2a/concepts-pr84979-3.C: Likewise.
* g++.dg/cpp2a/concepts-pr84979.C: Likewise.
* g++.dg/DRs/dr625.C: New test.

Replace free with XDELETE.

gcc/c/ChangeLog:

* c-typeck.c (free_all_tagged_tu_seen_up_to): Replace free
with XDELETE.
(finish_init): Likewise.
(pop_init_level): Likewise.

c++: Deducing type from initializer_list<auto> [PR93107]

In this testcase we weren't able to deduce b's type:

  template<typename T> void Task() { }
  auto b = { &Task<int> };

because resolve_nondeduced_context doesn't iterate on the {}'s elements.
So make sure to look into {} too.  We don't need to handle nested {}
here.

We could either tweak resolve_nondeduced_context to handle CONSTRUCTORs
or add a _ctor version, but then resolve_nondeduced_context_or_error
would need some changes too -- it'd have to check the result of a call
to r_n_c for each element.

gcc/cp/ChangeLog:

PR c++/93107
* pt.c (do_auto_deduction): Call resolve_nondeduced_context for
the elements of a { } list.

gcc/testsuite/ChangeLog:

PR c++/93107
* g++.dg/cpp0x/initlist-deduce3.C: New test.

c++: Reject float <=> enum.

As [depr.arith.conv.enum] says, these are ill-formed.

gcc/cp/ChangeLog:

* typeck.c (do_warn_enum_conversions): Don't warn for SPACESHIP_EXPR.
(cp_build_binary_op): Reject float <=> enum or enum <=> float.  Use
CP_INTEGRAL_TYPE_P instead of INTEGRAL_OR_ENUMERATION_TYPE_P.

gcc/testsuite/ChangeLog:

* g++.dg/cpp2a/enum-conv1.C: Remove unused code.
* g++.dg/cpp2a/spaceship-err5.C: New test.

c++: Simplify constraint normalization routines

Many of the high-level constraint normalization routines allow the
caller to supply the initial template arguments for normalization, but
in practice all of the callers supply something equivalent to the
identity mapping(*).

This patch hard-codes this prevalent choice of initial template
arguments by making get_normalized_constraints always pass NULL_TREE as
the args to normalize_expression.  This admits some simplifications in
the high-level routines, such as removing their 'args' parameter and
consolidating the two versions of normalize_constraint_expression.

(*): In particular, a set of generic template arguments or NULL_TREE.
In the case of the two-parm version of normalize_constraint_expression,
we were suspiciously using the template arguments of a concept-id when
normalizing the concept-id as a constraint-expression.

gcc/cp/ChangeLog:

* constraint.cc (get_normalized_constraints): Remove 'args'
parameter.  Pass NULL_TREE as the initial template arguments to
normalize_expression.
(get_normalized_constraints_from_info): Remove 'args' parameter
and adjust the call to get_normalized_constraints.
(get_normalized_constraints_from_decl): Remove 'args' local
variable and adjust call to get_normalized_constraints_from_info.
(normalize_concept_definition): Remove 'args' local variable
and adjust call to get_normalized_constraints.
(normalize_constraint_expression): Remove the two-parameter
overload.  Remove 'args' parameter from the three-parameter
overload and update function comment accordingly.  Remove
default argument from 'diag' parameter.  Adjust call to
get_normalized_constraints.
(finish_nested_requirement): Adjust call to
normalize_constraint_expression.
(strictly_subsumes): Remove 'args' parameter.  Adjust call to
get_normalized_constraints_from_info.
(weakly_subsumes): Likewise.
* cp-tree.h (strictly_subsumes): Remove 'args' parameter.
(weakly_subsumes): Likewise.
* pt.c (process_partial_specialization): Adjust call to
strictly_subsumes.
(is_compatible_template_arg): Adjust call to weakly_subsumes.

c++: Tolerate empty initial args during normalization [PR97412]

When normalizing the constraint-expression of a nested-requirement, we
pass NULL_TREE as the initial template arguments for normalization, but
tsubst_argument_pack is not prepared to handle a NULL_TREE args vector.
This causes us to ICE when normalizing a variadic concept as part of a
nested-requirement.

This patch fixes the ICE by guarding the call to tsubst_template_args in
normalize_concept_check appropriately.  This will also enable us to
simplify many of the normalization routines to just pass NULL_TREE
(instead of a set of generic template arguments) as the initial template
arguments.

gcc/cp/ChangeLog:

PR c++/97412
* constraint.cc (normalize_concept_check): Don't call
tsubst_template_args when 'args' is NULL.

gcc/testsuite/ChangeLog:

PR c++/97412
* g++.dg/cpp2a/concepts-variadic2.C: New test.

Fix some memleaks

This fixes some memleaks, one older, one recently introduced.

2020-10-29  Richard Biener  <rguenther@suse.de>

* tree-ssa-pre.c (compute_avail): Free operands consistently.
* tree-vect-loop.c (vectorizable_phi): Make sure all operand
defs vectors are released.

c++: Fix constexpr cleanup error handling.

In this testcase, the primary evaluation successfully produces 'true', and
then running one of the cleanups hits a double delete, making the whole
thing not a valid constant expression.  So we were returning 'true' wrapped
in a NOP_EXPR to indicate its non-constancy, but evaluating that again is a
perfectly acceptable constant expression, so we weren't getting the verbose
diagnostic we were looking for.

So if non_constant_p gets set other than for overflow, go back to the
original expression.

With this change, we should never hit the manifestly_const_eval test, and
the is-constant-evaluated1.C test passes without it.

gcc/cp/ChangeLog:

PR c++/97388
* constexpr.c (cxx_eval_outermost_constant_expr): Revert to
original expression if evaluation sets non_constant_p.

gcc/testsuite/ChangeLog:

PR c++/97388
* g++.dg/cpp2a/constexpr-dtor8.C: New test.

c++: Fix constexpr dtors vs invisible ref [PR97388]

For arguments passed by invisible reference, in the IL until genericization
we have the source types on the callee side and while on the caller side
we already pass references to the actual argument slot in the caller, we
undo that in cxx_bind_parameters_in_call's
      if (TREE_ADDRESSABLE (type))
        /* Undo convert_for_arg_passing work here.  */
        x = convert_from_reference (x);
This works fine most of the time, except when the type also has constexpr
destructor; in that case the destructor is invoked in the caller and thus
the unsharing we do to make sure that the callee doesn't modify caller's
values is in that case undesirable, it prevents the changes done in the
callee propagating to the caller which should see them for the constexpr
dtor evaluation.

The following patch fixes that.  While it could be perhaps done for all
TREE_ADDRESSABLE types, I don't see the need to change the behavior
if there is no constexpr non-trivial dtor.

Jason: And we need to avoid memoizing the call, because a later equivalent
call also needs to modify its argument.  And we don't need to unshare
constructors when we aren't memoizing the call, because we already unshared
them when evaluating the TARGET_EXPR representing the copy-initialization of
the argument.

2020-10-20  Jakub Jelinek  <jakub@redhat.com>
    Jason Merrill  <jason@redhat.com>

PR c++/97388
* constexpr.c (cxx_bind_parameters_in_call): Set non_constant_args
if the parameter type has a non-trivial destructor.
(cxx_eval_call_expression): Only unshare arguments if we're
memoizing this evaluation.

* g++.dg/cpp2a/constexpr-dtor5.C: New test.
* g++.dg/cpp2a/constexpr-dtor6.C: New test.
* g++.dg/cpp2a/constexpr-dtor7.C: New test.

Annotate vec::copy better for mem stats

* vec.h (vec<T, va_heap, vl_ptr>::copy): Pass mem stat info.

Avoid typeless storage in wide-int

* wide-int.h (trailing_wide_ints <N>): Turn len to array of structures
so it does not imply typeless storage.
(trailing_wide_ints <N>::operator): update
(trailing_wide_ints <N>::operator []): Update.

c++: Diagnose constexpr delete [] new int; and delete new int[N]; [PR95808]

This patch diagnoses delete [] new int; and delete new int[1]; in constexpr
contexts by remembering
IDENTIFIER_OVL_OP_FLAGS (DECL_NAME (fun)) & OVL_OP_FLAG_VEC
from the operator new and checking it at operator delete time.

2020-10-29  Jakub Jelinek  <jakub@redhat.com>

PR c++/95808
* cp-tree.h (enum cp_tree_index): Add CPTI_HEAP_VEC_UNINIT_IDENTIFIER
and CPTI_HEAP_VEC_IDENTIFIER.
(heap_vec_uninit_identifier, heap_vec_identifier): Define.
* decl.c (initialize_predefined_identifiers): Initialize those
identifiers.
* constexpr.c (cxx_eval_call_expression): Reject array allocations
deallocated with non-array deallocation or non-array allocations
deallocated with array deallocation.
(non_const_var_error): Handle heap_vec_uninit_identifier and
heap_vec_identifier too.
(cxx_eval_constant_expression): Handle also heap_vec_uninit_identifier
and in that case during initialization replace it with
heap_vec_identifier.
(find_heap_var_refs): Handle heap_vec_uninit_identifier and
heap_vec_identifier too.

* g++.dg/cpp2a/constexpr-new15.C: New test.

stdbool.h: Update true and false expansions for C2x

C2x has changed the expansions of the true and false macros in
<stdbool.h> so that they have type _Bool (including in #if conditions,
i.e. an unsigned type in that context).  Use the new expansions in
GCC's <stdbool.h> for C2x.

See bug 82272 for related discussion (but this patch does *not*
implement the warning discussed there).

Note that it's possible there may be a further change to make bool,
true and false keywords (there was support in principle for that at
the April WG14 meeting).  But currently these expansions of type _Bool
are what C2x requires and there isn't actually a paper before WG14 at
present that would introduce the new keywords.

Bootstrapped with no regressions on x86_64-pc-linux-gnu.

gcc/
2020-10-29  Joseph Myers  <joseph@codesourcery.com>

* ginclude/stdbool.h [__STDC_VERSION__ > 201710L] (true, false):
Define with type _Bool.

gcc/testsuite/
2020-10-29  Joseph Myers  <joseph@codesourcery.com>

* gcc.dg/c11-bool-1.c, gcc.dg/c2x-bool-1.c, gcc.dg/c99-bool-4.c:
New tests.

libstdc++: Improve tests for constexpr algorithms

These tests just return true without checking that the results of the
algorithms. Although it should be safe to assume that the algorithms
behave the same at compile-time as at run-time, we can use these tests
to verify it.

This replaces each 'return true' statement with a condition that depends
on the basic functionality of the algorithm, such as returning an
iterator to the right position.

libstdc++-v3/ChangeLog:

* testsuite/25_algorithms/all_of/constexpr.cc: Check result of
the algorithm.
* testsuite/25_algorithms/any_of/constexpr.cc: Likewise.
* testsuite/25_algorithms/binary_search/constexpr.cc: Likewise.
* testsuite/25_algorithms/copy_backward/constexpr.cc: Likewise.
* testsuite/25_algorithms/count/constexpr.cc: Likewise.
* testsuite/25_algorithms/equal/constexpr.cc: Likewise.
* testsuite/25_algorithms/equal_range/constexpr.cc: Likewise.
* testsuite/25_algorithms/fill/constexpr.cc: Likewise.
* testsuite/25_algorithms/find_end/constexpr.cc: Likewise.
* testsuite/25_algorithms/find_if/constexpr.cc: Likewise.
* testsuite/25_algorithms/is_partitioned/constexpr.cc: Likewise.
* testsuite/25_algorithms/is_permutation/constexpr.cc: Likewise.
* testsuite/25_algorithms/is_sorted_until/constexpr.cc:
Likewise.
* testsuite/25_algorithms/lexicographical_compare/constexpr.cc:
Likewise.
* testsuite/25_algorithms/lower_bound/constexpr.cc: Likewise.
* testsuite/25_algorithms/merge/constexpr.cc: Likewise.
* testsuite/25_algorithms/mismatch/constexpr.cc: Likewise.
* testsuite/25_algorithms/none_of/constexpr.cc: Likewise.
* testsuite/25_algorithms/partition_copy/constexpr.cc: Likewise.
* testsuite/25_algorithms/remove_copy/constexpr.cc: Likewise.
* testsuite/25_algorithms/remove_copy_if/constexpr.cc: Likewise.
* testsuite/25_algorithms/remove_if/constexpr.cc: Likewise.
* testsuite/25_algorithms/replace_if/constexpr.cc: Likewise.
* testsuite/25_algorithms/reverse/constexpr.cc: Likewise.
* testsuite/25_algorithms/reverse_copy/constexpr.cc: Likewise.
* testsuite/25_algorithms/rotate_copy/constexpr.cc: Likewise.
* testsuite/25_algorithms/search/constexpr.cc: Likewise.
* testsuite/25_algorithms/set_difference/constexpr.cc: Likewise.
* testsuite/25_algorithms/set_intersection/constexpr.cc:
Likewise.
* testsuite/25_algorithms/set_symmetric_difference/constexpr.cc:
Likewise.
* testsuite/25_algorithms/set_union/constexpr.cc: Likewise.
* testsuite/25_algorithms/unique_copy/constexpr.cc: Likewise.
* testsuite/25_algorithms/upper_bound/constexpr.cc: Likewise.

libstdc++: Allow Lemire's algorithm to be used in more cases

This extends the fast path to also work when the URBG's range of
possible values is not the entire range of its result_type. Previously,
the slow path would be used for engines with a uint_fast32_t result type
if that type is actually a typedef for uint64_t rather than uint32_t.
After this change, the generator's result_type is not important, only
the range of possible value that generator can produce. If the
generator's range is exactly UINT64_MAX then the calculation will be
done using 128-bit and 64-bit integers, and if the range is UINT32_MAX
it will be done using 64-bit and 32-bit integers.

In practice, this benefits most of the engines and engine adaptors
defined in [rand.predef] on x86_64-linux and other 64-bit targets. This
is because std::minstd_rand0 and std::mt19937 and others use
uint_fast32_t, which is a typedef for uint64_t.

The code now makes use of the recently-clarified requirement that the
generator's min() and max() functions are usable in constant
expressions (see LWG 2154).

libstdc++-v3/ChangeLog:

* include/bits/uniform_int_dist.h (_Power_of_two): Add
constexpr.
(uniform_int_distribution::_S_nd): Add static_assert to ensure
the wider type is twice as wide as the result type.
(uniform_int_distribution::__generate_impl): Add static_assert
and declare variables as constexpr where appropriate.
(uniform_int_distribution:operator()): Likewise. Only consider
the uniform random bit generator's range of possible results
when deciding whether _S_nd can be used, not the __uctype type.

libstdc++: Do not use volatile for __gnu_cxx::rope reference counting

The rope extension uses a volatile variable for its reference count.
This is not only unnecessary for correctness (volatile provides neither
atomicity nor memory visibility, and the variable is only modified while
a lock is held) but it now causes deprecated warnings with
-Wsystem-headers due to the use of ++ and -- operators.

It would be possible to use __gnu_cxx::__exchange_and_add in _M_incr and
_M_decr when __atomic_is_lock_free(sizeof(_RC_t), &_M_ref_count) is
true, rather than locking a mutex. That would probably be a significant
improvement for multi-threaded and single-threaded code (because
__exchange_and_add will use non-atomic ops when possible, and even in MT
code it should be faster than the mutex lock/unlock pair). However,
mixing objects compiled with the old and new code would result in
inconsistent synchronization being used for the reference count.

libstdc++-v3/ChangeLog:

* include/ext/rope (_Refcount_Base::_M_ref_count): Remove
volatile qualifier.
(_Refcount_Base::_M_decr()): Likewise.

libstdc++: Make std::function work better with -fno-rtti

This change allows std::function::target<F>() to work even without RTTI,
using the same approach as std::any. Because we know what the manager
function would be for a given type, we can check if the stored pointer
has the expected address. If it does, we don't need to use RTTI. If it
isn't equal, we still need to do the RTTI check (when RTTI is enabled)
to handle the case where the same function has different addresses in
different shared objects.

This also changes the implementation of the manager function to return a
null pointer result when asked for the type_info of the target object.
This not only avoids a warning with -Wswitch -Wsystem-headers, but also
avoids prevents std::function::target_type() from dereferencing an
uninitialized pointer when the linker keeps an instantiation of the
manager function that was compiled without RTTI.

Finally, this fixes a bug in the non-const overload of function::target
where calling it with a function type F was ill-formed, due to
attempting to use const_cast<F*>(ptr). The standard only allows
const_cast<T*> when T is an object type.  The solution is to use
*const_cast<F**>(&ptr) instead, because F* is an object type even if F
isn't. I've also used _GLIBCXX17_CONSTEXPR in function::target so that
it doesn't bother instantiating anything for types that can never be a
valid target.

libstdc++-v3/ChangeLog:

* include/bits/std_function.h (_Function_handler<void, void>):
Define explicit specialization used for invalid target types.
(_Base_manager::_M_manager) [!__cpp_rtti]: Return null.
(function::target_type()): Check for null pointer.
(function::target()): Define unconditionall. Fix bug with
const_cast of function pointer type.
(function::target() const): Define unconditionally, but
only use RTTI if enabled.
* testsuite/20_util/function/target_no_rtti.cc: New test.

Selectively trap if ranger and vr-values disagree on range builtins.

The UBSAN builtins degrade into PLUS/MINUS/MULT and call
extract_range_from_binary_expr, which as the PR shows, can special
case some symbolics which the ranger doesn't currently handle.

Looking at vr_values::extract_range_builtin(), I see that every single
place where we ask for a range, we bail on non-integers (symbolics,
etc).  That is, with the exception of the UBSAN builtins.

Since this seems to be particular to UBSAN, we could still go with the
original plan of removing the duplicity in ranger vs vr-values, but
leave in the UBSAN builtin handling.  This isn't ideal, as we'd like
to remove all the common code, but I'd be willing to put up with UBSAN
duplication for the time being.

This patch disables the assert on the UBSAN builtins, while still
trapping if any other differences are found between the vr_values and
the ranger versions of builtin range handling.

As a follow-up, once Fedora can test this approach, I'll remove all
the builtin code from extract_range_builtin, with the exception of the
UBSAN stuff (renaming it to extract_range_ubsan_builtin).

Since the builtin code has proven fickle across architectures, I've
tested this with {-m32,-m64,-fsanitize=signed-integer-overflow} on
x86, ppc64le, and aarch64.  I think this should be enough.  If it
isn't, we can revert the patch, and leave the duplicate code until
the next release cycle when hopefully vr_values, evrp, and friends
will all be overhauled.

gcc/ChangeLog:

PR tree-optimization/97505
* vr-values.c (vr_values::extract_range_basic): Enable
trap again for everything except UBSAN builtins.

* gcc.dg/pr97596.c: Require int128 effective target.

Add testcase for PR97596

Add testcase

gcc/testsuite
* gcc.dg/pr97596.c: New.

c++: Stop (most) function-scope entities having a template header

Currently push_template_decl (mostly) decides whether to add a
template header to an entity by seeing if it has DECL_LANG_SPECIFIC.
That might have been a useful predicate at one time, but basing
semantic implications on how we've decided to represent decls is bound
to be brittle.  And indeed it is, as more decls grow a use for
lang-specific.  In particular I discovered that function-scope
VAR_DECLs couild grow lang-specific, and thereby get a template
header.  There's no need for that, and it breaks an invariant modules
was expected.

This patch changes that, and bases the descision on the properties of
the decl.  In particular the only function-scope decl that gets a
template header is an implicit-typedef.

I also cleaned up the behaviour of it building a template-info only to
ignore it.

gcc/cp/
* pt.c (push_template_decl): Do not give function-scope entities
other than implicit typedefs a template header. Do not readd
template info to a redeclared template.

libstdc++: Fix memory issue in ranges::lexicographical_compare testcase

libstdc++-v3/ChangeLog:

* testsuite/25_algorithms/lexicographical_compare/constrained.cc:
(test03): Fix initializing the vector vy with the array y of size 4.

vect: Fix load costs for SLP permutes

For the following test case (compiled with load/store lanes
disabled locally):

  void
  f (uint32_t *restrict x, uint8_t *restrict y, int n)
  {
    for (int i = 0; i < n; ++i)
      {
x[i * 2] = x[i * 2] + y[i * 2];
x[i * 2 + 1] = x[i * 2 + 1] + y[i * 2];
      }
  }

we have a redundant no-op permute on the x[] load node:

   node 0x4472350 (max_nunits=8, refcnt=2)
          stmt 0 _5 = *_4;
          stmt 1 _13 = *_12;
          load permutation { 0 1 }

Then, when costing it, we pick a cost of 1, even though we need 4 copies
of the x[] load to match a single y[] load:

   ==> examining statement: _5 = *_4;
   Vectorizing an unaligned access.
   vect_model_load_cost: unaligned supported by hardware.
   vect_model_load_cost: inside_cost = 1, prologue_cost = 0 .

The problem is that the code only considers the permutation for
the first scalar iteration, rather than for all VF iterations.

This patch tries to fix that by making vect_transform_slp_perm_load
calculate the value instead.

gcc/
* tree-vectorizer.h (vect_transform_slp_perm_load): Take an
optional extra parameter.
* tree-vect-slp.c (vect_transform_slp_perm_load): Calculate
the number of loads as well as the number of permutes, taking
the counting loop from...
* tree-vect-stmts.c (vect_model_load_cost): ...here.  Use the
value computed by vect_transform_slp_perm_load for ncopies.

LTO: get_section: add new argument

gcc/ChangeLog:

PR lto/97508
* langhooks.c (lhd_begin_section): Call get_section with
not_existing = true.
* output.h (get_section): Add new argument.
* varasm.c (get_section): Fail when NOT_EXISTING is true
and a section already exists.
* ipa-cp.c (ipcp_write_summary): Remove.
(ipcp_read_summary): Likewise.
* ipa-fnsummary.c (ipa_fn_summary_read): Always read jump
functions summary.
(ipa_fn_summary_write): Always stream it.

libstdc++: Correct PR number in ChangeLog entry

Consistently pass the vector type for scalar SLP cost compute

This avoids randomly (based on whether the stmt is
SLP_TREE_REPRESENTATIVE and not a pattern stmt) passing a vector
type or NULL to the add_stmt_cost hook for scalar code cost
compute.  For example the x86 backend uses only the vector type to
decide on the scalar computation mode which makes costing off.

So the following explicitely passes the vector type and uses
SLP_TREE_VECTYPE for this purpose.

2020-10-29  Richard Biener  <rguenther@suse.de>

* tree-vect-slp.c (vect_bb_slp_scalar_cost): Pass
SLP_TREE_VECTYPE to record_stmt_cost.

libstdc++: Fix some warnings in headers

These are usually suppressed without -Wsystem-headers.

libstdc++-v3/ChangeLog:

* include/bits/hashtable_policy.h (_Local_iterator_base): Cast
value to avoid -Wsign-compare warnings.
* include/bits/regex.h (sub_match::_M_str): Avoid narrowing
conversion.
* include/bits/regex_compiler.tcc (_Compiler::_M_quantifier):
Initialize variable to avoid -Wmaybe-uninitialized warning.
* include/bits/shared_ptr_base.h (_Sp_counted_deleter::_Impl):
Reorder mem-initializer-list to avoid -Wreorder warning.
* include/bits/stl_tree.h (_Rb_tree_impl): Explicitly
initialize base class in copy constructor.
* include/debug/safe_iterator.h (_Safe_iterator): Likewise.
* include/ext/debug_allocator.h: Reorder mem-initializer-list
to avoid -Wreorder warning.
* include/ext/throw_allocator.h (throw_allocator_limit)
(throw_allocator_random): Add user-declared assignment operators
to avoid -Wdeprecated-copy warnings.

Fix the logic conditioning VX_ENTER/LEAVE_TLS_DTOR

Fix a basic #if/#ifdef confusion which leads to improper
choices in some configurations.

2020-10-28  Olivier Hainque  <hainque@adacore.com>

libgcc/
* config/gthr-vxworks-tls.c: Fix preprocessor logic
controlling the definition of VX_ENTER_TLS_DTOR and
VX_LEAVE_TLS_DTOR based on a version major check.

Fix conditional inclusion guard in gthr-vxworks-thread.c

This fixes the name of the macro used to condition the
inclusion of an actual implementation of some of the gthread
support services for VxWorks, to agree with the side
defining that macro based on tests against the targetted
VxWorks version major.

2020-10-28  Olivier Hainque  <hainque@adacore.com>

libgcc/
* config/gthr-vxworks-thread.c: Fix name of macro used
to condition the inclusion of an actual implementation.

opts: Sanity check for param names.

gcc/ChangeLog:

* optc-gen.awk: Check that params start with -param=.
* params.opt: Fix ipa-jump-function-lookups.

CSE conversions within sincos

On platforms in which Aux_[Real_Type] involves non-NOP conversions
(e.g., between single- and double-precision, or between short float
and float), the conversions before the calls are CSEd too late for
sincos to combine calls.

This patch enables the sincos pass to CSE type casts used as arguments
to eligible calls before looking for other calls using the same
operand.

for  gcc/ChangeLog

* tree-ssa-math-opts.c (sincos_stats): Add conv_removed.
(execute_cse_conv_1): New.
(execute_cse_sincos_1): Call it.  Fix return within
FOR_EACH_IMM_USE_STMT.
(pass_cse_sincos::execute): Report conv_inserted.

for  gcc/testsuite/ChangeLog

* gnat.dg/sin_cos.ads: New.
* gnat.dg/sin_cos.adb: New.
* gcc.dg/sin_cos.c: New.

libstdc++: Rename _UniformRandomNumberGenerator parameters

The paper P0346R1 renamed uniform random number generators to
uniform random bit generators, to describe their purpose more
accurately. This makes that same change in one of the relevant
files (but not the others).

libstdc++-v3/ChangeLog:

* include/bits/uniform_int_dist.h (uniform_int_distribution):
Rename _UniformRandomNumberGenerator template parameters to
_UniformRandomBitGenerator, as per P0346R1.

Enable GCC to support Intel Key Locker ISA

gcc/ChangeLog

2018-12-15  Xuepeng Guo  <xuepeng.guo@intel.com>
    Hongyu Wang  <hongyu.wang@intel.com>
    Hongtao Liu  <hongtao.liu@intel.com>

* common/config/i386/cpuinfo.h (get_available_features):
Detect KL, AESKLE and WIDEKL features.
* common/config/i386/i386-common.c
(OPTION_MASK_ISA_KL_SET): New.
(OPTION_MASK_ISA_WIDEKL_SET): Likewise.
(OPTION_MASK_ISA_KL_UNSET): Likewise.
(OPTION_MASK_ISA_WIDEKL_UNSET): Likewise.
(OPTION_MASK_ISA2_AVX2_UNSET): Likewise.
(OPTION_MASK_ISA2_AVX_UNSET): Likewise.
(OPTION_MASK_ISA2_SSE4_2_UNSET): Likewise.
(OPTION_MASK_ISA2_SSE4_1_UNSET): Likewise.
(OPTION_MASK_ISA2_SSE4_UNSET): Likewise.
(OPTION_MASK_ISA2_SSSE3_UNSET): Likewise.
(OPTION_MASK_ISA2_SSE3_UNSET): Likewise.
(OPTION_MASK_ISA2_SSE2_UNSET): Likewise.
(OPTION_MASK_ISA2_SSE_UNSET): Likewise.
(ix86_handle_option): Handle kl and widekl, add dependency chain
for KL and SSE2.
* common/config/i386/i386-cpuinfo.h (enum processor_features):
(FEATURE_KL, FEATURE_AESKLE, FEATURE_WIDEKL): New.
* common/config/i386/i386-isas.h: Add ISA_NAMES_TABLE_ENTRY
for KL, AESKLE and WIDEKL.
* config.gcc: Add keylockerintrin.h.
* doc/invoke.texi: Document new option -mkl and -mwidekl.
* doc/extend.texi: Document kl and widekl.
* config/i386/cpuid.h (bit_KL, bit_AESKLE, bit_WIDEKL): New.
* config/i386/i386-builtin-types.def ((UINT, UINT, V2DI, V2DI, PVOID),
(UINT, UINT, V2DI, PVOID), (VOID, V2DI, V2DI, V2DI, UINT),
(UINT8, PV2DI, V2DI, PCVOID), (UINT8, PV2DI, PCV2DI, PCVOID)): New
function types.
* config/i386/i386-builtin.def: Add
__builtin_ia32_loadiwkey,
__builtin_ia32_aesdec128kl_u8,
__builtin_ia32_aesdec256kl_u8,
__builtin_ia32_aesenc128kl_u8,
__builtin_ia32_aesenc256kl_u8,
__builtin_ia32_aesdecwide128kl_u8,
__builtin_ia32_aesdecwide256kl_u8,
__builtin_ia32_aesencwide128kl_u8,
__builtin_ia32_aesencwide256kl_u8,
__builtin_ia32_encodekey128_u32,
__builtin_ia32_encodekey256_u32.
* config/i386/i386-c.c (ix86_target_macros_internal): Handle
kl and widekl.
* config/i386/i386-options.c (isa2_opts): Add -mkl and -mwidekl.
(ix86_option_override_internal): Handle KL and WIDEKL.
(ix86_valid_target_attribute_inner_p): Add attribute for kl and widekl.
* config/i386/i386-expand.c
(ix86_expand_builtin): Expand Keylocker Builtins.
* config/i386/i386.h (TARGET_KL): New.
(TARGET_KL_P): Likewise.
(TARGET_WIDEKL): Likewise.
(TARGET_WIDEKL_P): Likewise.
(PTA_KL): Likewise.
(PTA_WIDEKL): Likewise.
(PTA_TIGERLAKE): Add PTA_KL, PTA_WIDEKL.
(PTA_ALDERLAKE): Likewise.
* config/i386/i386.opt: Add new option mkl and mwidekl.
* config/i386/keylockerintrin.h: New header file for Keylocker.
* config/i386/immintrin.h: Include keylockerintrin.h.
* config/i386/predicates.md (encodekey128_operation): New
predicate.
(encodekey256_operation): Likewise.
(aeswidekl_operation): Likewise.
* config/i386/sse.md (UNSPECV_LOADIWKEY): New.
(UNSPECV_AESDEC128KLU8): Likewise.
(UNSPECV_AESENC128KLU8): Likewise.
(UNSPECV_AESDEC256KLU8): Likewise.
(UNSPECV_AESENC256KLU8): Likewise.
(UNSPECV_AESDECWIDE128KLU8): Likewise.
(UNSPECV_AESENCWIDE128KLU8): Likewise.
(UNSPECV_AESDECWIDE256KLU8): Likewise.
(UNSPECV_AESENCWIDE256KLU8): Likewise.
(UNSPECV_ENCODEKEY128U32): Likewise.
(UNSPECV_ENCODEKEY256U32): Likewise.
(encodekey128u32): New expander.
(encodekey256u32): Likewise.
(aes<aeswideklvariant>u8): Likewise.
(loadiwkey): New insn pattern.
(*encodekey128u32): Likewise.
(*encodekey256u32): Likewise.
(aes<aesklvariant>u8): Likewise.
(*aes<aeswideklvariant>u8): Likewise.

gcc/testsuite/ChangeLog

* gcc.target/i386/keylocker-aesdec128kl.c: New test.
* gcc.target/i386/keylocker-aesdec256kl.c: Likewise.
* gcc.target/i386/keylocker-aesdecwide128kl.c: Likewise.
* gcc.target/i386/keylocker-aesdecwide256kl.c: Likewise.
* gcc.target/i386/keylocker-aesenc128kl.c: Likewise.
* gcc.target/i386/keylocker-aesencwide128kl.c: Likewise.
* gcc.target/i386/keylocker-aesencwide256kl.c: Likewise.
* gcc.target/i386/keylocker-encodekey128.c: Likewise.
* gcc.target/i386/keylocker-encodekey256.c: Likewise.
* gcc.target/i386/keylocker-loadiwkey.c: Likewise.
* g++.dg/other/i386-2.C: Add -mkl and -mwidekl.
* g++.dg/other/i386-3.C: Likewise.
* gcc.target/i386/sse-12.c: Likewise.
* gcc.target/i386/sse-13.c: Likewise.
* gcc.target/i386/sse-14.c: Likewise.
* gcc.target/i386/sse-22.c: Add kl and widekl.
* gcc.target/i386/sse-23.c: Likewise.
* gcc.target/i386/funcspec-56.inc: Add new target attribute test.

More BB vectorization tweaks

This tweaks the op build from splats to allow loads marked as not
vectorizable.  It also amends some dump prints with the address of
the SLP node or the instance to better be able to debug things.

2020-10-29  Richard Biener  <rguenther@suse.de>

* tree-vect-slp.c (vect_build_slp_tree_2): Allow splatting
not vectorizable loads.
(vect_build_slp_instance): Amend dumping with address.
(vect_slp_convert_to_external): Likewise.

* gcc.dg/vect/bb-slp-pr65935.c: Adjust.

Call infer_non_null() directly when checking for non-null.

Simply call infer_non_null directly and avoid uneccessary checks of
the statement being modified.

gcc/
PR tree-optimization/97609
* gimple-range-cache.cc (non_null_ref::process_name): Call
infer_nonnull_range directly instead of infer_value_range.
gcc/testsuite/
* g++.dg/pr97609.C: New.

libstdc++: Fix new basic_stringbuf constructor

libstdc++-v3/ChangeLog:

* include/std/sstream (basic_stringbuf(__string_type&&, openmode)):
Call _M_init_syncbuf to set up get/put areas. Also qualify
std::move.

libgo: handle linking to NetBSD's versioned symbols

On NetBSD, for backwards compatibility, various libc symbols are
renamed to a symbol with a version suffix. For example, this is the
(abbreviated) definition of sigaction:

    int sigaction(...) __asm__ ("__sigaction14")

This poses a challenge for libgo, which attempts to link sigaction by
way of an "//extern" comment:

    //extern sigaction
    func sigaction(...)

This results in a reference to the deprecated compatibility symbol
"sigaction", rather than the desired "__sigaction14" symbol.

This patch introduces a new "//extern-sysinfo" comment to handle this
situation. The new mklinknames.awk script scans a package for these
comments and outputs a "//go:linkname" directive that links the wrapper
to the correct versioned symbol, as determined by parsing the __asm__
annotation on the function's declaration in gen-sysinfo.go.

For now, only the following packages are scanned by mklinknames.awk:

    os
    os/user
    runtime
    syscall

gotools/:
* Makefile.am (check-runtime): Add runtime_linknames.go to
--extrafiles.
* Makefile.in: Regenerate.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/265125

Daily bump.

analyzer: more non-determinism fixes

gcc/analyzer/ChangeLog:
* program-state.cc (sm_state_map::on_liveness_change): Sort the
leaking svalues before calling on_state_leak.
(program_state::detect_leaks): Likewise when calling
on_svalue_leak.
* region-model-reachability.cc
(reachable_regions::mark_escaped_clusters): Likewise when
calling on_escaped_function.

analyzer: fix false leak diagnostic on offsets from malloc [PR97608]

gcc/analyzer/ChangeLog:
PR analyzer/97608
* region-model-reachability.cc (reachable_regions::handle_sval):
Operands of reachable reversible operations are reachable.

gcc/testsuite/ChangeLog:
PR analyzer/97608
* gcc.dg/analyzer/malloc-1.c (test_42d): New.
* gcc.dg/analyzer/pr97608.c: New test.

analyzer: move svalue and region decls to their own header files

gcc/ChangeLog:
* Makefile.in (ANALYZER_OBJS): Add analyzer/complexity.o.

gcc/analyzer/ChangeLog:
* analyzer.h (class state_machine): New forward decl.
(class logger): Likewise.
(class visitor): Likewise.
* complexity.cc: New file, taken from svalue.cc.
* complexity.h: New file, taken from region-model.h.
* region-model.h: Include "analyzer/svalue.h" and
"analyzer/region.h".  Move struct complexity to complexity.h.
Move svalue, its subclasses and supporting decls to svalue.h.
Move region, its subclasses and supporting decls to region.h.
* region.cc: Include "analyzer/region.h".
(symbolic_region::symbolic_region): Move here from region-model.h.
* region.h: New file, based on material from region-model.h.
* svalue.cc: Include "analyzer/svalue.h".
(complexity::complexity): Move to complexity.cc.
(complexity::from_pair): Likewise.
* svalue.h: New file, based on material from region-model.h.

analyzer: fix more pointer-printing in logs

gcc/analyzer/ChangeLog:
* program-state.cc (sm_state_map::print): Guard the printing of
the origin pointer with !flag_dump_noaddr.
* region.cc (string_region::dump_to_pp): Likewise for
m_string_cst.

libstdc++: Fix linker script

libstdc++-v3/ChangeLog:

* config/abi/pre/gnu.ver (GLIBCXX_3.4.29): Remove duplicate
patterns.
(CXXABI_1.3.13): Restore missing piece.

[RS6000] Don't be too clever with dg-do run and dg-do compile

Otherwise some versions of dejagnu go ahead and run the vsx tests
below when they should not.  To best cope with older dejagnu, put
"run" before "compile", the idea being that if the second dg-do always
wins then that won't cause fails.

The altivec tests also need -save-temps for the scan-assembler test to
occur when vms_hw.

* gcc.target/powerpc/vsx-load-element-extend-char.c: Put "dg-do run"
before "dg-do compile", and make them mutually exclusive.
* gcc.target/powerpc/vsx-load-element-extend-int.c: Likewise.
* gcc.target/powerpc/vsx-load-element-extend-longlong.c: Likewise.
* gcc.target/powerpc/vsx-load-element-extend-short.c: Likewise.
* gcc.target/powerpc/vsx-store-element-truncate-char.c: Likewise.
* gcc.target/powerpc/vsx-store-element-truncate-int.c: Likewise.
* gcc.target/powerpc/vsx-store-element-truncate-longlong.c: Likewise.
* gcc.target/powerpc/vsx-store-element-truncate-short.c: Likewise.
* gcc.target/powerpc/altivec-consts.c: Likewise, add -save-temps.
* gcc.target/powerpc/le-altivec-consts.c: Likewise.

[RS6000] float128-type-2.c unsupported

I noticed this test is unsupported on power10 when looking through
test logs.  There seems no reason why that should be the case, ie.
the target test was meant to be powerpc64*-*-linux*.  And that
simplifies down further.

* gcc.target/powerpc/float128-type-1.c: Simplify target test.
* gcc.target/powerpc/float128-type-2.c: Likewise.

Re: testsuite: Enable and adjust powerpc fold-vec-extract/insert testcases

git commit badeac77f552 changed expected number of addi instructions,
causing these fails on powerpc-linux.

gcc.target/powerpc/fold-vec-insert-int-p9.c: \\maddi\\M found 12 times
FAIL: gcc.target/powerpc/fold-vec-insert-int-p9.c scan-assembler-times \\maddi\\M 8
gcc.target/powerpc/fold-vec-extract-char.p9.c: addi found 6 times
FAIL: gcc.target/powerpc/fold-vec-extract-char.p9.c scan-assembler-times addi 3
gcc.target/powerpc/fold-vec-extract-int.p9.c: \\maddi\\M found 6 times
FAIL: gcc.target/powerpc/fold-vec-extract-int.p9.c scan-assembler-times \\maddi\\M 3
gcc.target/powerpc/fold-vec-extract-longlong.p7.c: \\maddi\\M found 6 times
FAIL: gcc.target/powerpc/fold-vec-extract-longlong.p7.c scan-assembler-times \\maddi\\M 4
gcc.target/powerpc/fold-vec-extract-longlong.p8.c: \\maddi\\M found 6 times
FAIL: gcc.target/powerpc/fold-vec-extract-longlong.p8.c scan-assembler-times \\maddi\\M 4
changed by badeac77f552

I'm not at all sure why we are counting addi.  On linux I see
eight in fold-vec-insert-int-p9.c tearing down the stack frame in
function epilogues, and four in
        addi 9,1,16
        lvewx 0,0,9
For aix you have the above four but with a -16 offset.  There are no
stack frames, and you have four addressing stack red-zone as
        addi 9,1,-64

fold-vec-extract-char.p9.c on linux just has epilogue addi, aix has
red-zone addressing.  The same for fold-vec-extract-int.p9.c,
fold-vec-extract-longlong.p7.c and fold-vec-extract-longlong.p8.c.

It seems silly to count addi in a function epilogue, and fragile to
count them in code.  So remove the ilp32 addi checks.

* gcc.target/powerpc/fold-vec-extract-char.p9.c: Don't check addi
count for ilp32.
* gcc.target/powerpc/fold-vec-extract-int.p9.c: Likewise.
* gcc.target/powerpc/fold-vec-extract-longlong.p7.c: Likewise.
* gcc.target/powerpc/fold-vec-extract-longlong.p8.c: Likewise.
* gcc.target/powerpc/fold-vec-insert-int-p9.c: Likewise.