Missed function specialization + partial devirtualization

v8:
 1. Rebase to master with Martin's static function (r280043) comments merge.
    Boostrap/testsuite/SPEC2017 tested pass on Power8-LE.
 2. TODO:
    2.1. C++ devirt for multiple speculative call targets.
    2.2. ipa-icf ipa_merge_profiles refine with COMDAT inline testcase.

This patch aims to fix PR69678 caused by PGO indirect call profiling
performance issues.
The bug that profiling data is never working was fixed by Martin's pull
back of topN patches, performance got GEOMEAN ~1% improvement(+24% for
511.povray_r specifically).
Still, currently the default profile only generates SINGLE indirect target
that called more than 75%.  This patch leverages MULTIPLE indirect
targets use in LTO-WPA and LTO-LTRANS stage, as a result, function
specialization, profiling, partial devirtualization, inlining and
cloning could be done successfully based on it.
Performance can get improved from 0.70 sec to 0.38 sec on simple tests.
Details are:
  1.  PGO with topn is enabled by default now, but only one indirect
  target edge will be generated in ipa-profile pass, so add variables to enable
  multiple speculative edges through passes, speculative_id will record the
  direct edge index bind to the indirect edge, indirect_call_targets length
  records how many direct edges owned by the indirect edge, postpone gimple_ic
  to ipa-profile like default as inline pass will decide whether it is benefit
  to transform indirect call.
  2.  Use speculative_id to track and search the reference node matched
  with the direct edge's callee for multiple targets.  Actually, it is the
  caller's responsibility to handle the direct edges mapped to same indirect
  edge.  speculative_call_info will return one of the direct edge specified,
  this will leverage current IPA edge process framework mostly.
  3.  Enable LTO WPA/LTRANS stage multiple indirect call targets analysis for
  profile full support in ipa passes and cgraph_edge functions.  speculative_id
  can be set by make_speculative id when multiple targets are binded to
  one indirect edge, and cloned if new edge is cloned.  speculative_id
  is streamed out and stream int by lto like lto_stmt_uid.
  4.  Create and duplicate all speculative direct edge's call summary
  in ipa-fnsummary.c with auto_vec.
  5.  Add 1 in module testcase and 2 cross module testcases.
  6.  Bootstrap and regression test passed on Power8-LE.  No function
  and performance regression for SPEC2017.

gcc/ChangeLog

2020-01-14  Xiong Hu Luo  <luoxhu@linux.ibm.com>

PR ipa/69678
* cgraph.c (symbol_table::create_edge): Init speculative_id and
target_prob.
(cgraph_edge::make_speculative): Add param for setting speculative_id
and target_prob.
(cgraph_edge::speculative_call_info): Update comments and find reference
by speculative_id for multiple indirect targets.
(cgraph_edge::resolve_speculation): Decrease the speculations
for indirect edge, drop it's speculative if not direct target
left. Update comments.
(cgraph_edge::redirect_call_stmt_to_callee): Likewise.
(cgraph_node::dump): Print num_speculative_call_targets.
(cgraph_node::verify_node): Don't report error if speculative
edge not include statement.
(cgraph_edge::num_speculative_call_targets_p): New function.
* cgraph.h (int common_target_id): Remove.
(int common_target_probability): Remove.
(num_speculative_call_targets): New variable.
(make_speculative): Add param for setting speculative_id.
(cgraph_edge::num_speculative_call_targets_p): New declare.
(target_prob): New variable.
(speculative_id): New variable.
* ipa-fnsummary.c (analyze_function_body): Create and duplicate
  call summaries for multiple speculative call targets.
* cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
* ipa-profile.c (struct speculative_call_target): New struct.
(class speculative_call_summary): New class.
(class speculative_call_summaries): New class.
(call_sums): New variable.
(ipa_profile_generate_summary): Generate indirect multiple targets summaries.
(ipa_profile_write_edge_summary): New function.
(ipa_profile_write_summary): Stream out indirect multiple targets summaries.
(ipa_profile_dump_all_summaries): New function.
(ipa_profile_read_edge_summary): New function.
(ipa_profile_read_summary_section): New function.
(ipa_profile_read_summary): Stream in indirect multiple targets summaries.
(ipa_profile): Generate num_speculative_call_targets from
profile summaries.
* ipa-ref.h (speculative_id): New variable.
* ipa-utils.c (ipa_merge_profiles): Update with target_prob.
* lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
common_target_probability.   Stream out speculative_id and
num_speculative_call_targets.
(input_edge): Likewise.
* predict.c (dump_prediction): Remove edges count assert to be
precise.
* symtab.c (symtab_node::create_reference): Init speculative_id.
(symtab_node::clone_references): Clone speculative_id.
(symtab_node::clone_referring): Clone speculative_id.
(symtab_node::clone_reference): Clone speculative_id.
(symtab_node::clear_stmts_in_references): Clear speculative_id.
* tree-inline.c (copy_bb): Duplicate all the speculative edges
if indirect call contains multiple speculative targets.
* value-prof.h  (check_ic_target): Remove.
* value-prof.c  (gimple_value_profile_transformations):
Use void function gimple_ic_transform.
* value-prof.c  (gimple_ic_transform): Handle topn case.
Fix comment typos.  Change it to a void function.

gcc/testsuite/ChangeLog

2020-01-14  Xiong Hu Luo  <luoxhu@linux.ibm.com>

PR ipa/69678
* gcc.dg/tree-prof/indir-call-prof-topn.c: New testcase.
* gcc.dg/tree-prof/crossmodule-indir-call-topn-1.c: New testcase.
* gcc.dg/tree-prof/crossmodule-indir-call-topn-1a.c: New testcase.
* gcc.dg/tree-prof/crossmodule-indir-call-topn-2.c: New testcase.
* lib/scandump.exp: Dump executable file name.
* lib/scanwpaipa.exp: New scan-pgo-wap-ipa-dump.

Daily bump.

Add initial octeontx2 support.

This adds octeontx2 naming.  It currently uses the cortexa57
cost model and schedule model until I submit this.  This is
more a place holder to get the naming of the cores in GCC 10.
I will submit the cost model in the next couple of days.

ChangeLog:
* config/aarch64/aarch64-cores.def (octeontx2): New define.
(octeontx2t98): New define.
(octeontx2t96): New define.
(octeontx2t93): New define.
(octeontx2f95): New define.
(octeontx2f95n): New define.
(octeontx2f95mm): New define.
* config/aarch64/aarch64-tune.md: Regenerate.
* doc/invoke.texi (-mcpu=): Document the new cpu types.

Replace update_web_docs_libstdcxx_svn with update_web_docs_libstdcxx_git

This patch replaces the update_web_docs_libstdcxx_svn script, that
updates online documentation from its sources in the GCC repository, run
once a day from cron, with update_web_docs_libstdcxx_git.

* update_web_docs_libstdcxx_git: New file.
* update_web_docs_libstdcxx_svn: Remove.
* crontab: Use update_web_docs_libstdcxx_git.

testsuite: remove unused dg-line directive

gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic-test-paths-2.c: Remove unused dg-line
directive.

MAINTAINERS: Update my email address.

Revert "contrib: Add in the default push rule which was overridden"

This reverts commit b60563a8bf18b5a49431b5913f32f5c5ff8319d9.

Doesn't work as expected.

contrib: Add in the default push rule which was overridden

When we add a push rule, the default rule gets removed, so add that in
explicitly.  This needs to come last since otherwise it would match
the custom redirecting rules we have for personal and vendor
sub-spaces.

I also noticed that the push rule for the vendor subspace still had
a force push default.  We don't want that so remove it.

* gcc-git-customization.sh: Add back the default rule that
is lost by adding a custom push rule.
* git-fetch-vendor.sh: Likewise, also remove '+' from push specs.

IPA: Avoid segfault in devirtualization_time_bonus (PR 93223)

2020-01-13  Martin Jambor  <mjambor@suse.cz>

PR ipa/93223
* ipa-cp.c (devirtualization_time_bonus): Check whether isummary is
NULL.

testsuite/
* g++.dg/ipa/pr93223.C: New test.

PR c++/80265 - constexpr __builtin_mem*.

The library has already worked around this issue, but I was curious about
why it wasn't working.  The answer: because we were passing &var to fold,
which doesn't know about the constexpr values hash table.  Fixed by passing
&"str" instead.

* constexpr.c (cxx_eval_builtin_function_call): Expose STRING_CST
to str/mem builtins.

PR c++/33799 - destroy return value if local cleanup throws.

This is a pretty rare situation since the C++11 change to make all
destructors default to noexcept, but it is still possible to define throwing
destructors, and if a destructor for a local variable throws during the
return, we've already constructed the return value, so now we need to
destroy it.  I handled this somewhat like the new-expression cleanup; as in
that case, this cleanup can't properly nest with the cleanups for local
variables, so I introduce a cleanup region around the whole function and a
flag variable to indicate whether the return value actually needs to be
destroyed.

Setting the flag requires giving a COMPOUND_EXPR as the operand of a
RETURN_EXPR, so I adjust gimplify_return_expr to handle that.

This doesn't currently work with deduced return type because we don't know
the type when we're deciding whether to introduce the cleanup region.

gcc/
* gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
gcc/cp/
* cp-tree.h (current_retval_sentinel): New macro.
* decl.c (start_preparsed_function): Set up cleanup for retval.
* typeck.c (check_return_expr): Set current_retval_sentinel.

PR c++/93238 - short right-shift with enum.

My earlier patch to reintroduce short_shift failed to adjust for the C++
front end use of const_op1 rather than op1.

* typeck.c (cp_build_binary_op): Use folded op1 for short_shift.

Replace update_web_docs_svn with update_web_docs_git.

This patch replaces the update_web_docs_svn script, that updates
online documentation from its sources in the GCC repository, run once
a day from cron, with update_web_docs_git.

* update_web_docs_git: New file.
* update_web_docs_svn: Remove.
* crontab: Use update_web_docs_svn.

Fix typo in invoke.texi.

Fix handling of overflow in C casts in integer constant expressions (PR c/93241).

Bug 93241 reports a case where certain C expressions involving casts,
that would not be valid in an evaluated part of an integer constant
expression (because of e.g. involving integer overflow), are wrongly
rejected in an unevaluated part of an integer constant expression even
though all the operands and operations are ones that are valid in that
context.  This is a rejects-valid regression in GCC 4.5 and later
relative to 4.4 (for some testcases; the one in the bug uses
_Static_assert which isn't supported in those older releases).

The rule in the C front end is that an expression with those
properties (valid in an unevaluated part of an integer constant
expression but not an evaluated part) must be represented either as an
INTEGER_CST with TREE_OVERFLOW set or as a C_MAYBE_CONST_EXPR with
C_MAYBE_CONST_EXPR_INT_OPERANDS set.  This patch fixes build_c_cast to
check for that case and call note_integer_operands as needed.

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

PR c/93241
gcc/c:
* c-typeck.c (build_c_cast): Check for expressions with integer
operands that can occur in an unevaluated part of an integer
constant expression and call note_integer_operands as needed.

gcc/testsuite:
* gcc.dg/c11-static-assert-10.c, gcc.dg/c99-const-expr-15.c: New
tests.

contrib: script to setup git to pull a vendors branches

This simple script is intended to setup a new git configuration to
pull the branches and tags for a specific vendor.  This should
simplify some of the steps needed for working with a vendor's
branches.

* git-fetch-vendor.sh: New file.

contrib: Add git gcc-descr and gcc-undescr aliases.

2020-01-13  Jakub Jelinek  <jakub@redhat.com>

* contrib/gcc-git-customization.sh: Add git gcc-descr and gcc-undescr
aliases.

contrib: Add script to help with customizing a git checkout for use with GCC

This patch is intended to help with folks setting up a git work
environment for use with GCC following the transition to git.  It
currently does a couple of things.

1) Add an alias 'svn-rev' to git so that you can look up a legacy
commit by its svn revision number.  This enables you to type
git svn-rev 1234
and git will show the commit log entry relating to SVN r1234.

2) Sets up tracking information for the user's personal area in
the git repo.  It tries to figure out some sensible answers to the
data it needs, but allows the user to override the values.  It then
creates the fetch and push entries that are needed for tracking the
extra refs. This implements one part of the recommendations that I've
proposed in svnwrite.html for dealing with private branches.

It should be possible to run the script more than once and for it to
DTRT.  If you change your answers the configuration should be
correctly updated.

2020-01-13  Richard Earnshaw  <rearnsha@arm.com>

    * gcc-git-customization: New file.

Remove usage of legacy param_ipa_cp_unit_growth.

* ipa-cp.c (get_max_overall_size): Use newly
renamed param param_ipa_cp_unit_growth.
* params.opt: Remove legacy param name.

PR tree-optimization/93213 - wrong code with -Og -foptimize-strlen

gcc/testsuite/ChangeLog:

PR tree-optimization/93213
* gcc.c-torture/execute/pr93213.c: New test.

gcc/ChangeLog:

PR tree-optimization/93213
* tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
stores to be eliminated.

Exclude CL_WARNING and CL_PARAM from --help=optimize.

2020-01-13  Martin Liska  <mliska@suse.cz>

* opts.c (print_help): Do not print CL_PARAM
and CL_WARNING for CL_OPTIMIZATION.

libstdc++: Ensure root-dir converted to forward slash (PR93244)

PR libstdc++/93244
* include/bits/fs_path.h (path::generic_string<C,A>)
[_GLIBCXX_FILESYSTEM_IS_WINDOWS]: Convert root-dir to forward-slash.
* testsuite/27_io/filesystem/path/generic/generic_string.cc: Check
root-dir is converted to forward slash in generic pathname.
* testsuite/27_io/filesystem/path/generic/utf.cc: New test.
* testsuite/27_io/filesystem/path/generic/wchar_t.cc: New test.

libstdc++: Value-initialize std::atomic for C++20 (P0883R2)

This implements the new requirements for C++20 that std::atomic should
initialize the atomic variable in its default constructor.

This patch does not add the deprecated attribute to atomic_init, but
that should be done at some point as it's deprecated in C++20.

The paper also deprecates the ATOMIC_FLAG_INIT macro, although we can't
apply the deprecated attribute to a macro.

PR libstdc++/58605
* include/bits/atomic_base.h (__cpp_lib_atomic_value_initialization):
Define.
(__atomic_flag_base, __atomic_base, __atomic_base<_PTp*>)
(__atomic_float): Add default member initializer for C++20.
* include/std/atomic (atomic): Likewise.
(atomic::atomic()): Remove noexcept-specifier on default constructor.
* include/std/version (__cpp_lib_atomic_value_initialization): Define.
* testsuite/29_atomics/atomic/cons/assign_neg.cc: Adjust dg-error line
number.
* testsuite/29_atomics/atomic/cons/copy_neg.cc: Likewise.
* testsuite/29_atomics/atomic/cons/value_init.cc: New test.
* testsuite/29_atomics/atomic_flag/cons/value_init.cc: New test.
* testsuite/29_atomics/atomic_flag/requirements/trivial.cc: Adjust
expected result for is_trivially_default_constructible.
* testsuite/29_atomics/atomic_float/requirements.cc: Likewise.
* testsuite/29_atomics/atomic_float/value_init.cc: New test.
* testsuite/29_atomics/atomic_integral/cons/assign_neg.cc: Likewise.
* testsuite/29_atomics/atomic_integral/cons/copy_neg.cc: Likewise.
* testsuite/29_atomics/atomic_integral/cons/value_init.cc
* testsuite/29_atomics/atomic_integral/requirements/trivial.cc: Adjust
expected results for is_trivially_default_constructible.
* testsuite/util/testsuite_common_types.h (has_trivial_dtor): Add
new test generator.

doc: Note that some warnings depend on optimizations (PR 92757)

PR driver/92757
* doc/invoke.texi (Warning Options): Add caveat about some warnings
depending on optimization settings.

tree-opt: Fix bootstrap failure in tree-ssa-forwprop.c some more PR90838

2020-01-13  Jakub Jelinek  <jakub@redhat.com>

PR tree-optimization/90838
* tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
argument rather than to initialize temporary for targets that
don't use the mode argument at all.  Initialize ctzval to avoid
warning at -O0.

Update gcc_release for move to git.

This patch updates the gcc_release script, used for snapshots and
releases, to use git instead of SVN.

The parts of the changes used in building snapshots have been tested.
The parts used for building releases have not, so it's likely further
fixes may turn out to be needed there when releases are first built
from git.

* gcc_release: Use git instead of SVN.
* crontab: Update gcc_release calls.

Daily bump.

Replace update_version_svn with update_version_git.

This patch replaces the update_version_svn script, that updates the
DATESTAMP files from cron, with update_version_git.

* update_version_git: New file.
* update_version:svn: Remove.
* crontab: Use update_version_git.

Remove README.MOVED_TO_GIT.  The repository is now open for commits.

* README.MOVED_TO_GIT: Remove.

Add README.MOVED_TO_GIT.

* README.MOVED_TO_GIT: New file.

From-SVN: r280156

Daily bump.

From-SVN: r280155

[amdgcn] Remove dependency on stdint.h in libgcc

2020-01-10  Kwok Cheung Yeung  <kcy@codesourcery.com>

libgcc/
* config/gcn/atomic.c: Remove include of stdint.h.
(__sync_val_compare_and_swap_##SIZE): Replace uintptr_t with
__UINTPTR_TYPE__.

From-SVN: r280152

OpenACC 'acc_get_property' cleanup

include/
* gomp-constants.h (enum gomp_device_property): Remove.
libgomp/
* libgomp-plugin.h (enum goacc_property): New.  Adjust all users
to use this instead of 'enum gomp_device_property'.
(GOMP_OFFLOAD_get_property): Rename to...
(GOMP_OFFLOAD_openacc_get_property): ... this.  Adjust all users.
* libgomp.h (struct gomp_device_descr): Move
'GOMP_OFFLOAD_openacc_get_property'...
(struct acc_dispatch_t): ... here.  Adjust all users.
* plugin/plugin-hsa.c (GOMP_OFFLOAD_get_property): Remove.
liboffloadmic/
* plugin/libgomp-plugin-intelmic.cpp (GOMP_OFFLOAD_get_property):
Remove.

From-SVN: r280150

Further changes for the OpenACC 'if_present' clause on the 'host_data' construct

gcc/
* tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
* tree-core.h: Document it.
* gimplify.c (gimplify_omp_workshare): Set it.
* omp-low.c (lower_omp_target): Use it.
* tree-pretty-print.c (dump_omp_clause): Print it.
gcc/testsuite/
* c-c++-common/goacc/host_data-1.c: Extend.
* gfortran.dg/goacc/host_data-tree.f95: Likewise.

gcc/
* omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.

libgomp/
* target.c (gomp_map_vars_internal)
<GOMP_MAP_USE_DEVICE_PTR_IF_PRESENT>: Clean up/elaborate code
paths.

From-SVN: r280149

libstdc++: Improve comment about testsuite utilities

This fixes a typo and also explains why test_container is not a range
when used with output_iterator_wrapper or input_iterator_wrapper.

* testsuite/util/testsuite_iterators.h: Improve comment.

From-SVN: r280146

Add diagnostic paths

This patch adds support for associating a "diagnostic_path" with a
diagnostic: a sequence of events predicted by the compiler that leads to
the problem occurring, with their locations in the user's source,
text descriptions, and stack information (for handling interprocedural
paths).

For example, the following (hypothetical) error has a 3-event
intraprocedural path:

test.c: In function 'demo':
test.c:29:5: error: passing NULL as argument 1 to 'PyList_Append' which
  requires a non-NULL parameter
   29 |     PyList_Append(list, item);
      |     ^~~~~~~~~~~~~~~~~~~~~~~~~
  'demo': events 1-3
     |
     |   25 |   list = PyList_New(0);
     |      |          ^~~~~~~~~~~~~
     |      |          |
     |      |          (1) when 'PyList_New' fails, returning NULL
     |   26 |
     |   27 |   for (i = 0; i < count; i++) {
     |      |   ~~~
     |      |   |
     |      |   (2) when 'i < count'
     |   28 |     item = PyLong_FromLong(random());
     |   29 |     PyList_Append(list, item);
     |      |     ~~~~~~~~~~~~~~~~~~~~~~~~~
     |      |     |
     |      |     (3) when calling 'PyList_Append', passing NULL from (1) as argument 1
     |

The patch adds a new "%@" format code for printing event IDs, so that
in the above, the description of event (3) mentions event (1), showing
the user where the bogus NULL value comes from (the event IDs are
colorized to draw the user's attention to them).

There is a separation between data vs presentation: the above shows how
the diagnostic-printing code has consolidated the path into a single run
of events, since all the events are near each other and within the same
function; more complicated examples (such as interprocedural paths)
might be printed as multiple runs of events.

Examples of how interprocedural paths are printed can be seen in the
test suite (which uses a plugin to exercise the code without relying
on specific warnings using this functionality).

Other output formats include
- JSON,
- printing each event as a separate "note", and
- to not emit paths.

gcc/ChangeLog:
* Makefile.in (OBJS): Add tree-diagnostic-path.o.
* common.opt (fdiagnostics-path-format=): New option.
(diagnostic_path_format): New enum.
(fdiagnostics-show-path-depths): New option.
* coretypes.h (diagnostic_event_id_t): New forward decl.
* diagnostic-color.c (color_dict): Add "path".
* diagnostic-event-id.h: New file.
* diagnostic-format-json.cc (json_from_expanded_location): Make
non-static.
(json_end_diagnostic): Call context->make_json_for_path if it
exists and the diagnostic has a path.
(diagnostic_output_format_init): Clear context->print_path.
* diagnostic-path.h: New file.
* diagnostic-show-locus.c (colorizer::set_range): Special-case
when printing a run of events in a diagnostic_path so that they
all get the same color.
(layout::m_diagnostic_path_p): New field.
(layout::layout): Initialize it.
(layout::print_any_labels): Don't colorize the label text for an
event in a diagnostic_path.
(gcc_rich_location::add_location_if_nearby): Add
"restrict_to_current_line_spans" and "label" params.  Pass the
former to layout.maybe_add_location_range; pass the latter
when calling add_range.
* diagnostic.c: Include "diagnostic-path.h".
(diagnostic_initialize): Initialize context->path_format and
context->show_path_depths.
(diagnostic_show_any_path): New function.
(diagnostic_path::interprocedural_p): New function.
(diagnostic_report_diagnostic): Call diagnostic_show_any_path.
(simple_diagnostic_path::num_events): New function.
(simple_diagnostic_path::get_event): New function.
(simple_diagnostic_path::add_event): New function.
(simple_diagnostic_event::simple_diagnostic_event): New ctor.
(simple_diagnostic_event::~simple_diagnostic_event): New dtor.
(debug): New overload taking a diagnostic_path *.
* diagnostic.def (DK_DIAGNOSTIC_PATH): New.
* diagnostic.h (enum diagnostic_path_format): New enum.
(json::value): New forward decl.
(diagnostic_context::path_format): New field.
(diagnostic_context::show_path_depths): New field.
(diagnostic_context::print_path): New callback field.
(diagnostic_context::make_json_for_path): New callback field.
(diagnostic_show_any_path): New decl.
(json_from_expanded_location): New decl.
* doc/invoke.texi (-fdiagnostics-path-format=): New option.
(-fdiagnostics-show-path-depths): New option.
(-fdiagnostics-color): Add "path" to description of default
GCC_COLORS; describe it.
(-fdiagnostics-format=json): Document how diagnostic paths are
represented in the JSON output format.
* gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
Add optional params "restrict_to_current_line_spans" and "label".
* opts.c (common_handle_option): Handle
OPT_fdiagnostics_path_format_ and
OPT_fdiagnostics_show_path_depths.
* pretty-print.c: Include "diagnostic-event-id.h".
(pp_format): Implement "%@" format code for printing
diagnostic_event_id_t *.
(selftest::test_pp_format): Add tests for "%@".
* selftest-run-tests.c (selftest::run_tests): Call
selftest::tree_diagnostic_path_cc_tests.
* selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
* toplev.c (general_init): Initialize global_dc->path_format and
global_dc->show_path_depths.
* tree-diagnostic-path.cc: New file.
* tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
non-static.  Drop "diagnostic" param in favor of storing the
original value of "where" and re-using it.
(virt_loc_aware_diagnostic_finalizer): Update for dropped param of
maybe_unwind_expanded_macro_loc.
(tree_diagnostics_defaults): Initialize context->print_path and
context->make_json_for_path.
* tree-diagnostic.h (default_tree_diagnostic_path_printer): New
decl.
(default_tree_make_json_for_path): New decl.
(maybe_unwind_expanded_macro_loc): New decl.

gcc/c-family/ChangeLog:
* c-format.c (local_event_ptr_node): New.
(PP_FORMAT_CHAR_TABLE): Add entry for "%@".
(init_dynamic_diag_info): Initialize local_event_ptr_node.
* c-format.h (T_EVENT_PTR): New define.

gcc/testsuite/ChangeLog:
* gcc.dg/format/gcc_diag-10.c (diagnostic_event_id_t): New
typedef.
(test_diag): Add coverage of "%@".
* gcc.dg/plugin/diagnostic-path-format-default.c: New test.
* gcc.dg/plugin/diagnostic-path-format-inline-events-1.c: New test.
* gcc.dg/plugin/diagnostic-path-format-inline-events-2.c: New test.
* gcc.dg/plugin/diagnostic-path-format-inline-events-3.c: New test.
* gcc.dg/plugin/diagnostic-path-format-none.c: New test.
* gcc.dg/plugin/diagnostic-test-paths-1.c: New test.
* gcc.dg/plugin/diagnostic-test-paths-2.c: New test.
* gcc.dg/plugin/diagnostic-test-paths-3.c: New test.
* gcc.dg/plugin/diagnostic-test-paths-4.c: New test.
* gcc.dg/plugin/diagnostic_plugin_test_paths.c: New.
* gcc.dg/plugin/plugin.exp: Add the new plugin and test cases.

libcpp/ChangeLog:
* include/line-map.h (class diagnostic_path): New forward decl.
(rich_location::get_path): New accessor.
(rich_location::set_path): New function.
(rich_location::m_path): New field.
* line-map.c (rich_location::rich_location): Initialize m_path.

From-SVN: r280142

re PR tree-optimization/93210 (Sub-optimal code optimization on struct/combound constexpr (gcc vs. clang))

PR tree-optimization/93210
* fold-const.h (native_encode_initializer,
can_native_interpret_type_p): Declare.
* fold-const.c (native_encode_string): Fix up handling with off != -1,
simplify.
(native_encode_initializer): New function, moved from dwarf2out.c.
Adjust to native_encode_expr compatible arguments, including dry-run
and partial extraction modes.  Don't handle STRING_CST.
(can_native_interpret_type_p): No longer static.
* gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
offset / BITS_PER_UNIT fits into int and don't call it if
can_native_interpret_type_p fails.  If suboff is NULL and for
CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
native_encode_initializer.
(fold_const_aggregate_ref_1): Formatting fix.
* dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
(tree_add_const_value_attribute): Adjust caller.

* gcc.dg/pr93210.c: New test.
* g++.dg/opt/pr93210.C: New test.

From-SVN: r280141

re PR tree-optimization/90838 (Detect table-based ctz implementation)

PR tree-optimization/90838
* tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
CTZ_DEFINED_VALUE_AT_ZERO.

From-SVN: r280140

re PR inline-asm/93027 (ICE: in match_reload, at lra-constraints.c:1060)

2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>

PR inline-asm/93027
* gcc.target/i386/pr93027.c: Use the right PR number in the test.

From-SVN: r280138

re PR libgomp/93219 (unused return value in affinity-fmt.c)

PR libgomp/93219
* libgomp.h (gomp_print_string): Change return type from void to int.
* affinity-fmt.c (gomp_print_string): Likewise.  Return true if
not all characters have been written.

From-SVN: r280137

Rename pr93207.c to pr93027.c.  Fix PR number in change logs.

From-SVN: r280136

re PR inline-asm/93027 (ICE: in match_reload, at lra-constraints.c:1060)

2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>

PR inline-asm/93207
* gcc.target/i386/pr93207.c: Run it only for x86-64.

From-SVN: r280135

testsuite: add dg-enable-nn-line-numbers

This patch adds support for obscuring the line numbers printed in the
left-hand margin when printing the source code, converting them to "NN",
e.g from:

  7111 |   if (!(flags & 0x0001)) {
       |      ^
       |      |
       |      (1) following 'true' branch...
  7112 |

to:

   NN  |   if (!(flags & 0x0001)) {
       |      ^
       |      |
       |      (1) following 'true' branch...
   NN  |

This is useful in followup patches e.g. when testing how interprocedural
paths are printed using multiline.exp, to avoid depending on precise line
numbers.

gcc/testsuite/ChangeLog:
* lib/gcc-dg.exp (cleanup-after-saved-dg-test): Reset global
nn_line_numbers_enabled.
* lib/multiline.exp (nn_line_numbers_enabled): New global.
(dg-enable-nn-line-numbers): New proc.
(maybe-handle-nn-line-numbers): New proc.
* lib/prune.exp (prune_gcc_output): Call maybe-handle-nn-line-numbers.

From-SVN: r280134

re PR inline-asm/93027 (ICE: in match_reload, at lra-constraints.c:1060)

2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>

PR inline-asm/93207
* lra-constraints.c (match_reload): Permit input operands have the
same mode as output while other input operands have a different
mode.

2020-01-10  Vladimir Makarov  <vmakarov@redhat.com>

PR inline-asm/93207
* gcc.target/i386/pr93207.c: New test.

From-SVN: r280133

PR90838: Support ctz idioms

Support common idioms for count trailing zeroes using an array lookup.
The canonical form is array[((x & -x) * C) >> SHIFT] where C is a magic
constant which when multiplied by a power of 2 creates a unique value
in the top 5 or 6 bits.  This is then indexed into a table which maps it
to the number of trailing zeroes.  When the table is valid, we emit a
sequence using the target defined value for ctz (0):

int ctz1 (unsigned x)
{
  static const char table[32] =
    {
      0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8,
      31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9
    };

  return table[((unsigned)((x & -x) * 0x077CB531U)) >> 27];
}

Is optimized to:

rbit w0, w0
clz w0, w0
and w0, w0, 31
ret

    gcc/
PR tree-optimization/90838
* tree-ssa-forwprop.c (check_ctz_array): Add new function.
(check_ctz_string): Likewise.
(optimize_count_trailing_zeroes): Likewise.
(simplify_count_trailing_zeroes): Likewise.
(pass_forwprop::execute): Try ctz simplification.
* match.pd: Add matching for ctz idioms.

    testsuite/
PR tree-optimization/90838
* testsuite/gcc.target/aarch64/pr90838.c: New test.

From-SVN: r280132

aarch64.c (aarch64_invalid_conversion): New function for target hook.

gcc/ChangeLog:

2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>

* config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
for target hook.
(aarch64_invalid_unary_op): New function for target hook.
(aarch64_invalid_binary_op): New function for target hook.

gcc/testsuite/ChangeLog:

2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>

* g++.target/aarch64/bfloat_cpp_typecheck.C: New test.
* gcc.target/aarch64/bfloat16_scalar_typecheck.c: New test.
* gcc.target/aarch64/bfloat16_vector_typecheck_1.c: New test.
* gcc.target/aarch64/bfloat16_vector_typecheck_2.c: New test.

From-SVN: r280130

config.gcc: Add arm_bf16.h.

2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>

* config.gcc: Add arm_bf16.h.
* config/aarch64/aarch64-builtins.c
(aarch64_simd_builtin_std_type): Add BFmode.
(aarch64_init_simd_builtin_types): Define element types for vector
types.
(aarch64_init_bf16_types): New function.
(aarch64_general_init_builtins): Add arm_init_bf16_types function call.
* config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
modes.
* config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
* config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
patterns.
* config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
(AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
* config/aarch64/aarch64.c
(aarch64_classify_vector_mode): Add support for BF types.
(aarch64_gimplify_va_arg_expr): Add support for BF types.
(aarch64_vq_mode): Add support for BF types.
(aarch64_simd_container_mode): Add support for BF types.
(aarch64_mangle_type): Add support for BF scalar type.
* config/aarch64/aarch64.md: Add BFmode to movhf pattern.
* config/aarch64/arm_bf16.h: New file.
* config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
* config/aarch64/iterators.md: Add BF types to mode attributes.
(HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.

2020-01-10  Stam Markianos-Wright  <stam.markianos-wright@arm.com>

* g++.dg/abi/mangle-neon-aarch64.C: Add Bfloat SIMD types to test.
* g++.dg/ext/arm-bf16/bf16-mangle-aarch64-1.C: New test.
* gcc.target/aarch64/bfloat16_scalar_1.c: New test.
* gcc.target/aarch64/bfloat16_scalar_2.c: New test.
* gcc.target/aarch64/bfloat16_scalar_3.c: New test.
* gcc.target/aarch64/bfloat16_scalar_4.c: New test.
* gcc.target/aarch64/bfloat16_simd_1.c: New test.
* gcc.target/aarch64/bfloat16_simd_2.c: New test.
* gcc.target/aarch64/bfloat16_simd_3.c: New test.

From-SVN: r280129

Shorten right-shift again in C++.

Back in SVN r131862 richi removed this code to fix PR 34235, but didn't
remove the parallel code from the C front-end because the bug had previously
been fixed in r44080.  This patch copies the code from C again.

* typeck.c (cp_build_binary_op): Restore short_shift code.

From-SVN: r280128

PR c++/93143 - incorrect tree sharing with constexpr.

We don't unshare CONSTRUCTORs as often during constexpr evaluation, so we
need to unshare them here.

* constexpr.c (cxx_eval_outermost_constant_expr): Don't assume
CONSTRUCTORs are already unshared.

From-SVN: r280127

PR c++/93173 - incorrect tree sharing.

My patch for 93033 wasn't sufficient to handle all the possible sharing
introduced by split_nonconstant_init, and it occurred to me that it would
make sense to use the same unsharing technique as unshare_body, namely
copy_if_shared.

PR c++/93033
gcc/
* gimplify.c (copy_if_shared): No longer static.
* gimplify.h: Declare it.
gcc/cp/
* cp-gimplify.c (cp_gimplify_init_expr, cp_gimplify_expr): Use
copy_if_shared after cp_genericize_tree.
* typeck2.c (split_nonconstant_init): Don't unshare here.

From-SVN: r280126

[AArch64] Make -msve-vector-bits=128 generate VL-specific code

related_vector_mode and compatible_vector_types_p make it possible
to generate 128-bit SVE code while still maintaining the distinction
between SVE vectors and Advanced SIMD vectors.  We can therefore generate
VL-specific code for -msve-vector-bits=128 on little-endian targets.
In theory we could do the same for big-endian targets, but it could
have quite a high overhead; see the comment in the patch for details.

2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* doc/invoke.texi (-msve-vector-bits=): Document that
-msve-vector-bits=128 now generates VL-specific code for
little-endian targets.
* config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
build_vector_type_for_mode to construct the data vector types.
* config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
VL-specific code for -msve-vector-bits=128 on little-endian targets.
(aarch64_simd_container_mode): Always prefer Advanced SIMD modes
for 128-bit vectors.

gcc/testsuite/
* gcc.target/aarch64/sve/struct_vect_1.c (N): Protect with #ifndef.
* gcc.target/aarch64/sve/pcs/return_1_128.c: New test.
* gcc.target/aarch64/sve/pcs/return_4_128.c: Likewise.
* gcc.target/aarch64/sve/pcs/return_5_128.c: Likewise.
* gcc.target/aarch64/sve/pcs/return_6_128.c: Likewise.
* gcc.target/aarch64/sve/pcs/stack_clash_1_128.c: Likewise.
* gcc.target/aarch64/sve/pcs/stack_clash_2_128.c: Likewise.
* gcc.target/aarch64/sve/single_5.c: Likewise.
* gcc.target/aarch64/sve/struct_vect_25.c: Likewise.
* gcc.target/aarch64/sve/struct_vect_26.c: Likewise.

From-SVN: r280125

PR c/93132 - bogus 'attribute((access))' warning when size-index is specified

gcc/c-family/ChangeLog:

PR c/93132
* c-attribs.c (append_access_attrs): Validate against the translated
access string rather than the human-readable representation.

gcc/testsuite/ChangeLog:

PR c/93132
* gcc.dg/attr-access-read-only-2.c: New test.

From-SVN: r280124

[AArch64] Fix reversed vcond_mask invocation in aarch64_evpc_sel

aarch64_evpc_sel (new in GCC 10) got the true and false vectors
the wrong way round, leading to execution failures with fixed-length
128-bit SVE.

Now that the ACLE types are in trunk, it's much easier to match
the exact asm sequence for a permute.

2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
invocation.

gcc/testsuite/
* gcc.target/aarch64/sve/sel_1.c: Use SVE types for the arguments and
return values.  Use check-function-bodies instead of scan-assembler.
* gcc.target/aarch64/sve/sel_2.c: Likewise
* gcc.target/aarch64/sve/sel_3.c: Likewise.

From-SVN: r280121

gotest: don't use local

    It's not part of the POSIX shell standard.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/214300

From-SVN: r280118

libstdc++: Fix testcase for C++98 compatibility

* testsuite/25_algorithms/equal/deque_iterators/1.cc: Don't use C++11
initialization syntax.

From-SVN: r280117

libstdc++: Make istreambuf_iterator base class consistent (PR92285)

Since LWG 445 was implemented for GCC 4.7, the std::iterator base class
of std::istreambuf_iterator changes type depending on the -std mode
used. This creates an ABI incompatibility between different -std modes.

This change ensures the base class always has the same type. This makes
layout for C++98 compatible with the current -std=gnu++14 default, but
no longer compatible with C++98 code from previous releases. In practice
this is unlikely to cause real problems, because it only affects the
layout of types with two std::iterator base classes, one of which comes
from std::istreambuf_iterator. Such types are expected to be vanishingly
rare.

PR libstdc++/92285
* include/bits/streambuf_iterator.h (istreambuf_iterator): Make type
of base class independent of __cplusplus value.
[__cplusplus < 201103L] (istreambuf_iterator::reference): Override the
type defined in the base class
* testsuite/24_iterators/istreambuf_iterator/92285.cc: New test.
* testsuite/24_iterators/istreambuf_iterator/requirements/
base_classes.cc: Adjust expected base class for C++98.

From-SVN: r280116

OpenACC – support "if" + "if_present" clauses with "host_data"

2020-01-10  Gergö Barany  <gergo@codesourcery.com>
    Thomas Schwinge <thomas@codesourcery.com>
    Julian Brown  <julian@codesourcery.com>
    Tobias Burnus  <tobias@codesourcery.com>

        gcc/c/
        * c-parser.c (OACC_HOST_DATA_CLAUSE_MASK): Add PRAGMA_OACC_CLAUSE_IF
        and PRAGMA_OACC_CLAUSE_IF_PRESENT.

        gcc/cp/
        * parser.c (OACC_HOST_DATA_CLAUSE_MASK): Add PRAGMA_OACC_CLAUSE_IF
        and PRAGMA_OACC_CLAUSE_IF_PRESENT.

        gcc/fortran/
        * openmp.c (OACC_HOST_DATA_CLAUSES): Add PRAGMA_OACC_CLAUSE_IF
        and PRAGMA_OACC_CLAUSE_IF_PRESENT.

gcc/
* omp-low.c (lower_omp_target): Use GOMP_MAP_USE_DEVICE_PTR_IF_PRESENT
if PRAGMA_OACC_CLAUSE_IF_PRESENT exist.

gcc/testsuite/
* c-c++-common/goacc/host_data-1.c: Added tests of if and if_present
clauses on host_data.
* gfortran.dg/goacc/host_data-tree.f95: Likewise.

include/
* gomp-constants.h (enum gomp_map_kind): New enumeration constant
GOMP_MAP_USE_DEVICE_PTR_IF_PRESENT.

libgomp/
* oacc-parallel.c (GOACC_data_start): Handle
GOMP_MAP_USE_DEVICE_PTR_IF_PRESENT.
* target.c (gomp_map_vars_async): Likewise.
* testsuite/libgomp.oacc-c-c++-common/host_data-7.c: New.
* testsuite/libgomp.oacc-fortran/host_data-5.F90: New.

From-SVN: r280115

[AArch64] Tighten mode checks in aarch64_builtin_vectorized_function

aarch64_builtin_vectorized_function checked vectors based on the
number of elements and the element mode.  This doesn't interact
well with fixed-length 128-bit SVE, where SVE modes can have those
same properties.  (And we can't just use the built-ins for SVE because
the types use a different ABI.  SVE handles this kind of thing using
optabs instead.)

2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/aarch64-builtins.c
(aarch64_builtin_vectorized_function): Check for specific vector modes,
rather than checking the number of elements and the element mode.

From-SVN: r280114

Use get_related_vectype_for_scalar_type for reduction indices

The related_vector_mode series missed this case in
vect_create_epilog_for_reduction, where we want to create the
unsigned integer equivalent of another vector.  Without it we
could mix SVE and Advanced SIMD vectors in the same operation.

This showed up on existing tests when testing with fixed-length
-msve-vector-bits=128.

2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* tree-vect-loop.c (vect_create_epilog_for_reduction): Use
get_related_vectype_for_scalar_type rather than build_vector_type
to create the index type for a conditional reduction.

From-SVN: r280112

Fix gather/scatter check when updating a vector epilogue loop

update_epilogue_loop_vinfo applies SSA renmaing to the DR_REF of a
gather or scatter, so that vect_check_gather_scatter continues to work.
However, we sometimes also rely on vect_check_gather_scatter when
using gathers and scatters to implement strided accesses.

This showed up on existing tests when testing with fixed-length
-msve-vector-bits=128.

2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
for any type of gather or scatter, including strided accesses.

From-SVN: r280111

compiler: permit duplicate methods from embedded interfaces

    This is a language change for Go 1.14.

    Updates golang/go#6977

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/214240

From-SVN: r280109

[vect] Add missing comment

gcc/ChangeLog:
2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>

* tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
 comment.

From-SVN: r280108

[vect] Keep track of DR_OFFSET advance in dr_vec_info rather than data_reference

gcc/ChangeLog:
2020-01-10  Andre Vieira  <andre.simoesdiasvieira@arm.com>

* tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
get_dr_vinfo_offset
* tree-vect-loop.c (update_epilogue_loop_vinfo):  Remove orig_drs_init
parameter and its use to reset DR_OFFSET's.
(vect_transform_loop): Remove orig_drs_init argument.
* tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
member of dr_vec_info rather than the offset of the associated
data_reference's innermost_loop_behavior.
(vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
(vect_do_peeling): Remove orig_drs_init parameter and its construction.
* tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
get_dr_vinfo_offset.
(vectorizable_store): Likewise.
(vectorizable_load): Likewise.

From-SVN: r280107

2020-01-10  Richard Biener  <rguenther@suse.de>

* gimple-ssa-store-merging
(pass_store_merging::terminate_all_aliasing_chains): Cache alias info.

From-SVN: r280106

Fix ipa-clone-3.c on some targets

2020-01-10  Martin Jambor  <mjambor@suse.cz>

* gcc.dg/ipa/ipa-clone-3.c: Replace struct initializer with
piecemeal initialization.

From-SVN: r280105

[AArch64] Require aarch64_sve256_hw for a 256-bit SVE test

One of the SVE run tests was specific to 256-bit SVE but tried to
run for all SVE lengths.

2020-01-10  Richard Sandiford  <richard.sandiford@arm.com>

gcc/testsuite/
* gcc.target/aarch64/sve/index_1_run.c: Require aarch64_sve256_hw
rather than aarch64_sve_hw.

From-SVN: r280104

Fix wrong parenthesis in inliner.

2020-01-10  Martin Liska  <mliska@suse.cz>

PR ipa/93217
* ipa-inline-analysis.c (offline_size): Make proper parenthesis
encapsulation that was there before r280040.

From-SVN: r280103

re PR tree-optimization/93199 (Compile time hog in sink_clobbers)

2020-01-10  Richard Biener  <rguenther@suse.de>

PR middle-end/93199
* tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
sequences to avoid walking them again for secondary opportunities.
(pass_lower_eh_dispatch::execute): Instead actually insert
them here.

From-SVN: r280102

re PR tree-optimization/93199 (Compile time hog in sink_clobbers)

2020-01-10  Richard Biener  <rguenther@suse.de>

PR middle-end/93199
* tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
(cleanup_all_empty_eh): Walk landing pads in reverse order to
avoid quadraticness.

From-SVN: r280101

IPA-CP: Access param_ipa_sra_max_replacements through opt_for_fn

2020-01-10  Martin Jambor  <mjambor@suse.cz>

* params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
* ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
to get param_ipa_sra_max_replacements.
(param_splitting_across_edge): Pass the caller to
pull_accesses_from_callee.

From-SVN: r280100

IPA-CP: Always access param_ipcp_unit_growth through opt_for_fn

2020-01-10  Martin Jambor  <mjambor@suse.cz>

* params.opt (param_ipcp_unit_growth): Mark as Optimization.
* ipa-cp.c (max_new_size): Removed.
(orig_overall_size): New variable.
(get_max_overall_size): New function.
(estimate_local_effects): Use it.  Adjust dump.
(decide_about_value): Likewise.
(ipcp_propagate_stage): Do not calculate max_new_size, just store
orig_overall_size.  Adjust dump.
(ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.

From-SVN: r280099

IPA-CP: Always access param_ipa_max_agg_items through opt_for_fn

2020-01-10  Martin Jambor  <mjambor@suse.cz>

* params.opt (param_ipa_max_agg_items): Mark as Optimization
* ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
instead of param_ipa_max_agg_items.
(merge_aggregate_lattices): Extract param_ipa_max_agg_items from
optimization info for the callee.

From-SVN: r280098

re PR testsuite/93216 (gcc.dg/optimize-bswaphi-1.c fails starting with r280034)

2020-01-10  Richard Biener  <rguenther@suse.de>

PR testsuite/93216
* gcc.dg/optimize-bswaphi-1.c: Split previously added
case into a LE and BE variant.

From-SVN: r280097

Daily bump.

From-SVN: r280096

libgo: compile examples in _test packages

    Previously if the only names defined by _test packages were examples,
    the gotest script would emit an incorrect _testmain.go file.
    I worked around that by marking the example_test.go files +build ignored.

    This CL changes the gotest script to handle this case correctly,
    and removes the now-unnecessary build tags.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/214039

From-SVN: r280085

rename local _C2 identifiers in stl map header files

2020-01-09  Olivier Hainque  <hainque@adacore.com>

* doc/xml/manual/appendix_contributing.xml: Document _C2
as a reserved identifier, by VxWorks.
* include/bits/stl_map.h: Rename _C2 template typenames as _Cmp2.
* include/bits/stl_multimap.h: Likewise.

From-SVN: r280076

libstdc++: Fix <ext/pointer.h> incompatibilities with C++20

The equality operators for _ExtPtr_allocator are defined as non-const
member functions, which causes ambiguities in C++20 due to the
synthesized operator!= candidates. They should always have been const.

The _Pointer_adapter class template has both value_type and element_type
members, which makes readable_traits<_Pointer_adapter<T>> ambiguous. The
intended workaround is to add a specialization of readable_traits.

* include/ext/extptr_allocator.h (_ExtPtr_allocator::operator==)
(_ExtPtr_allocator::operator!=): Add missing const qualifiers.
* include/ext/pointer.h (readable_traits<_Pointer_adapter<S>>): Add
partial specialization to disambiguate the two constrained
specializations.

From-SVN: r280067

libstdc++: Fix testsuite failures and warnings due to is_pod deprecation

With -std=gnu++2a and -Wsystem-headers the std::is_pod deprecation
causes some new diagnostics. This suppresses them.

* include/experimental/type_traits (experimental::is_pod_v): Disable
-Wdeprecated-declarations warnings around reference to std::is_pod.
* include/std/type_traits (is_pod_v): Likewise.
* testsuite/18_support/max_align_t/requirements/2.cc: Also check
is_standard_layout and is_trivial. Do not check is_pod for C++20.
* testsuite/20_util/is_pod/requirements/explicit_instantiation.cc:
Add -Wno-deprecated for C++20.
* testsuite/20_util/is_pod/requirements/typedefs.cc: Likewise.
* testsuite/20_util/is_pod/value.cc: Likewise.
* testsuite/experimental/type_traits/value.cc: Likewise.

From-SVN: r280066

libstdc++: Implementing P0767 - deprecate POD

This adds the deprecated attribute to std::is_pod and std::is_pod_v for
C++20.

2019-12-05  JeanHeyd "ThePhD" Meneide  <phdofthehouse@gmail.com>

* include/bits/c++config (_GLIBCXX20_DEPRECATED): Add new macro.
* include/std/type_traits (is_pod, is_pod_v): Deprecate for C++20.
* testuite/20_util/is_pod/deprecated-2a.cc: New test.

From-SVN: r280065

libstdc++: Fix whitespace in ChangeLog-2019

From-SVN: r280064

Save typespec for empty array constructor.

2020-01-09  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/65428
* array.c (empty_constructor): New variable.
(empty_ts): New variable.
(expand_constructor): Save typespec in empty_ts.
Unset empty_constructor if there is an element.
(gfc_expand_constructor): Initialize empty_constructor
and empty_ts.  If there was no explicit constructor
type and the constructor is empty, take the type from
empty_ts.

2020-01-09  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/65428
* gfortran.dg/zero_sized_11.f90: New test.

From-SVN: r280063

Remove inline debug markers if support not enabled on accelerator compiler

2020-01-09  Kwok Cheung Yeung  <kcy@codesourcery.com>

gcc/
* lto-streamer-in.c (input_function): Remove streamed-in inline debug
markers if debug_inline_points is false.

From-SVN: r280062

libstdc++: Fix undefined behaviour in random dist serialization (PR93205)

The deserialization functions for random number distributions fail to
check the stream state before using the extracted values. In some cases
this leads to using indeterminate values to resize a vector, and then
filling that vector with indeterminate values.

No values that affect control flow should be used without checking that a
good value was read from the stream.

Additionally, where reasonable to do so, defer modifying any state in
the distribution until all values have been successfully read, to avoid
modifying some of the distribution's parameters and leaving others
unchanged.

PR libstdc++/93205
* include/bits/random.h (operator>>): Check stream operation succeeds.
* include/bits/random.tcc (operator<<): Remove redundant __ostream_type
typedefs.
(operator>>): Remove redundant __istream_type typedefs. Check stream
operations succeed.
(__extract_params): New function to fill a vector from a stream.
* testsuite/26_numerics/random/pr60037-neg.cc: Adjust dg-error line.

From-SVN: r280061

[AArch64] Add support for the SVE2 ACLE

This patch adds support for the SVE2 ACLE,  The implementation
and tests follow the same pattern as the exiting SVE ACLE support.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
extra_objs.
* config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
aarch64-sve-builtins-sve2.h.
(aarch64-sve-builtins-sve2.o): New rule.
* config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
(AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
(AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
(TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
* config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
TARGET_SVE2_SM4.
* config/aarch64/aarch64-sve.md: Update comments with SVE2
instructions that are handled here.
(@cond_asrd<mode>): Generalize to...
(@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
(*cond_asrd<mode>_2): Generalize to...
(*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
(*cond_asrd<mode>_z): Generalize to...
(*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
* config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
(UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
(UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
* config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
pattern.
(@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
(@aarch64_scatter_stnt<mode>): Likewise.
(@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
(@aarch64_mul_lane_<mode>): Likewise.
(@aarch64_sve_suqadd<mode>_const): Likewise.
(*<sur>h<addsub><mode>): Generalize to...
(@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
new pattern.
(@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
(*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
(*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
(*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
(*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
(@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
(@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
(@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
(@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
(*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
(*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
(*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
(@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
(@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
(@aarch64_sve_add_mul_lane_<mode>): Likewise.
(@aarch64_sve_sub_mul_lane_<mode>): Likewise.
(@aarch64_sve2_xar<mode>): Likewise.
(@aarch64_sve2_bcax<mode>): Likewise.
(*aarch64_sve2_eor3<mode>): Rename to...
(@aarch64_sve2_eor3<mode>): ...this.
(@aarch64_sve2_bsl<mode>): New expander.
(@aarch64_sve2_nbsl<mode>): Likewise.
(@aarch64_sve2_bsl1n<mode>): Likewise.
(@aarch64_sve2_bsl2n<mode>): Likewise.
(@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
(*aarch64_sve2_sra<mode>): Add MOVPRFX support.
(@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
(@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
(@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
(*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
(@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
(<su>mull<bt><Vwide>): Generalize to...
(@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
pattern.
(@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
(@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
(@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
(@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
(@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
(@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
(@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
(@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
(@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
(@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
(@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
(@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
(@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
(@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
(@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
(@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
(<SHRNB:r>shrnb<mode>): Generalize to...
(@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
new pattern.
(<SHRNT:r>shrnt<mode>): Generalize to...
(@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
new pattern.
(@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
(@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
(@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
(*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
(*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
(@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
(@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
(@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
(@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
(@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
(@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
(@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
(*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
(@aarch64_sve2_cvtnt<mode>): Likewise.
(@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
(@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
(*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
(@aarch64_sve2_cvtxnt<mode>): Likewise.
(@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
(@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
(*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
(@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
(@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
(*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
(@aarch64_sve2_pmul<mode>): Likewise.
(@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
(@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
(@aarch64_sve2_tbl2<mode>): Likewise.
(@aarch64_sve2_tbx<mode>): Likewise.
(@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
(@aarch64_sve2_histcnt<mode>): Likewise.
(@aarch64_sve2_histseg<mode>): Likewise.
(@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
(*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
(*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
(aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
(aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
(*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
(aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
(<su>mulh<r>s<mode>3): Update after above pattern name changes.
* config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
(SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
(SVE2_PMULL_PAIR_I): New mode iterators.
(UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
(UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
(UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
(UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
(UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
(UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
(UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
(UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
(UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
(UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
(UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
(UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
(UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
(UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
(UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
(UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
(UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
(UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
(UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
(UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
(UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
(UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
(UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
(UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
(UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
(UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
(UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
(UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
(UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
(UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
(UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
further down file.
(VNARROW, Ventype): New mode attributes.
(Vewtype): Handle VNx2DI.  Fix typo in comment.
(VDOUBLE): New mode attribute.
(sve_lane_con): Handle VNx8HI.
(SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
(SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
(sve_int_op, sve_int_op_rev): Handle the above codes.
(sve_pred_int_rhs2_operand): Likewise.
(MULLBT, SHRNB, SHRNT): Delete.
(SVE_INT_SHIFT_IMM): New int iterator.
(SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
and UNSPEC_WHILEHS for TARGET_SVE2.
(SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
(SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
(SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
(SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
(SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
(SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
(SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
(SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
(SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
(SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
(SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
(SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
(SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
(SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
(SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
(SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
(SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
(optab): Handle the new unspecs.
(su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
and UNSPEC_RSHRNT.
(lr): Handle the new unspecs.
(bt): Delete.
(cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
(sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
(sve_int_qsub_op): New int attributes.
(sve_fp_op, rot): Handle the new unspecs.
* config/aarch64/aarch64-sve-builtins.h
(function_resolver::require_matching_pointer_type): Declare.
(function_resolver::resolve_unary): Add an optional boolean argument.
(function_resolver::finish_opt_n_resolution): Add an optional
type_suffix_index argument.
(gimple_folder::redirect_call): Declare.
(gimple_expander::prepare_gather_address_operands): Add an optional
bool parameter.
* config/aarch64/aarch64-sve-builtins.cc: Include
aarch64-sve-builtins-sve2.h.
(TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
(TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
(TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
(TYPES_hsd_integer): Use TYPES_hsd_signed.
(TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
(TYPES_s_unsigned): Likewise.
(TYPES_s_integer): Use TYPES_s_unsigned.
(TYPES_sd_signed, TYPES_sd_unsigned): New macros.
(TYPES_sd_integer): Use them.
(TYPES_d_unsigned): New macro.
(TYPES_d_integer): Use it.
(TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
(TYPES_cvt_narrow): Likewise.
(DEF_SVE_TYPES_ARRAY): Include the new types macros above.
(preds_mx): New variable.
(function_builder::add_overloaded_function): Allow the new feature
set to be more restrictive than the original one.
(function_resolver::infer_pointer_type): Remove qualifiers from
the pointer type before printing it.
(function_resolver::require_matching_pointer_type): New function.
(function_resolver::resolve_sv_displacement): Handle functions
that don't support 32-bit vector indices or svint32_t vector offsets.
(function_resolver::finish_opt_n_resolution): Take the inferred type
as a separate argument.
(function_resolver::resolve_unary): Optionally treat all forms in
the same way as normal merging functions.
(gimple_folder::redirect_call): New function.
(function_expander::prepare_gather_address_operands): Add an argument
that says whether scaled forms are available.  If they aren't,
handle scaling of vector indices and don't add the extension and
scaling operands.
(function_expander::map_to_unspecs): If aarch64_sve isn't available,
fall back to using cond_* instead.
* config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
Split out the member variables into...
(rtx_code_function_base): ...this new base class.
(rtx_code_function_rotated): Inherit rtx_code_function_base.
(unspec_based_function): Split out the member variables into...
(unspec_based_function_base): ...this new base class.
(unspec_based_function_rotated): Inherit unspec_based_function_base.
(unspec_based_function_exact_insn): New class.
(unspec_based_add_function, unspec_based_add_lane_function)
(unspec_based_lane_function, unspec_based_pred_function)
(unspec_based_qadd_function, unspec_based_qadd_lane_function)
(unspec_based_qsub_function, unspec_based_qsub_lane_function)
(unspec_based_sub_function, unspec_based_sub_lane_function): New
typedefs.
(unspec_based_fused_function): New class.
(unspec_based_mla_function, unspec_based_mls_function): New typedefs.
(unspec_based_fused_lane_function): New class.
(unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
typedefs.
(CODE_FOR_MODE1): New macro.
(fixed_insn_function): New class.
(while_comparison): Likewise.
* config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
(binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
(binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
(load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
(load_gather_sv_restricted, shift_left_imm_long): Declare.
(shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
(shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
(shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
(store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
(ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
(ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
(unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
(unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
* config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
Also add an initial argument for unary_convert_narrowt, regardless
of the predication type.
(build_32_64): Allow loads and stores to specify MODE_none.
(build_sv_index64, build_sv_uint_offset): New functions.
(long_type_suffix): New function.
(binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
(binary_imm_long_base, load_gather_sv_base): Likewise.
(shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
(ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
(unary_narrowb_base, unary_narrowt_base): Likewise.
(binary_long_lane_def, binary_long_lane): New shape.
(binary_long_opt_n_def, binary_long_opt_n): Likewise.
(binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
(binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
(binary_to_uint_def, binary_to_uint): Likewise.
(binary_wide_def, binary_wide): Likewise.
(binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
(compare_def, compare): Likewise.
(compare_ptr_def, compare_ptr): Likewise.
(load_ext_gather_index_restricted_def,
load_ext_gather_index_restricted): Likewise.
(load_ext_gather_offset_restricted_def,
load_ext_gather_offset_restricted): Likewise.
(load_gather_sv_def): Inherit from load_gather_sv_base.
(load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
(shift_left_imm_def, shift_left_imm): Likewise.
(shift_left_imm_long_def, shift_left_imm_long): Likewise.
(shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
(store_scatter_index_restricted_def,
store_scatter_index_restricted): Likewise.
(store_scatter_offset_restricted_def,
store_scatter_offset_restricted): Likewise.
(tbl_tuple_def, tbl_tuple): Likewise.
(ternary_long_lane_def, ternary_long_lane): Likewise.
(ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
(ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
(ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
(ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
(ternary_qq_rotate_def, ternary_qq_rotate): New shape.
(ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
(ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
(ternary_uint_def, ternary_uint): Likewise.
(unary_convert): Fix typo in comment.
(unary_convert_narrowt_def, unary_convert_narrowt): New shape.
(unary_long_def, unary_long): Likewise.
(unary_narrowb_def, unary_narrowb): Likewise.
(unary_narrowt_def, unary_narrowt): Likewise.
(unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
(unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
(unary_to_int_def, unary_to_int): Likewise.
* config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
(unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
(svasrd_impl): Delete.
(svcadd_impl::expand): Handle integer operations too.
(svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
new functions to derive the unspec numbers.
(svmla_svmls_lane_impl): Replace with...
(svmla_lane_impl, svmls_lane_impl): ...these new classes.  Handle
integer operations too.
(svwhile_impl): Rename to...
(svwhilelx_impl): ...this and inherit from while_comparison.
(svasrd): Use unspec_based_function.
(svmla_lane): Use svmla_lane_impl.
(svmls_lane): Use svmls_lane_impl.
(svrecpe, svrsqrte): Handle unsigned integer operations too.
(svwhilele, svwhilelt): Use svwhilelx_impl.
* config/aarch64/aarch64-sve-builtins-sve2.h: New file.
* config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
* config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
* config/aarch64/aarch64-sve-builtins.def: Include
aarch64-sve-builtins-sve2.def.

gcc/testsuite/
* g++.target/aarch64/sve/acle/general-c++/mul_lane_1.c: New test.
* g++.target/aarch64/sve2/acle: New directory.
* gcc.target/aarch64/pragma_cpp_predefs_3.c: New test.
* gcc.target/aarch64/sve/acle/asm/test_sve_acle.h (TEST_TYPE_CHANGE_Z)
(TEST_DUAL_ZD, TEST_TYPE_CHANGE_ZX, TEST_TBL2, TEST_TBL2_REV): New
macros.
* gcc.target/aarch64/sve/acle/general-c/binary_lane_1.c: Do not
expect an error saying that the function has no f32 form, but instead
expect an error about SVE2 being required if the current target
doesn't support SVE2.
* gcc.target/aarch64/sve/acle/general-c/ternary_lane_1.c: Likewise.
* gcc.target/aarch64/sve/acle/general-c/ternary_lane_rotate_1.c Likewise.
* gcc.target/aarch64/sve/acle/general-c/binary_long_lane_1.c,
* gcc.target/aarch64/sve/acle/general-c/binary_long_opt_n_1.c,
* gcc.target/aarch64/sve/acle/general-c/binary_narrowb_opt_n_1.c,
* gcc.target/aarch64/sve/acle/general-c/binary_narrowt_opt_n_1.c,
* gcc.target/aarch64/sve/acle/general-c/binary_to_uint_1.c,
* gcc.target/aarch64/sve/acle/general-c/binary_wide_1.c,
* gcc.target/aarch64/sve/acle/general-c/binary_wide_opt_n_1.c,
* gcc.target/aarch64/sve/acle/general-c/compare_1.c,
* gcc.target/aarch64/sve/acle/general-c/compare_ptr_1.c,
* gcc.target/aarch64/sve/acle/general-c/load_ext_gather_index_restricted_1.c,
* gcc.target/aarch64/sve/acle/general-c/load_ext_gather_offset_restricted_1.c,
* gcc.target/aarch64/sve/acle/general-c/load_ext_gather_offset_restricted_2.c,
* gcc.target/aarch64/sve/acle/general-c/load_ext_gather_offset_restricted_3.c,
* gcc.target/aarch64/sve/acle/general-c/load_ext_gather_offset_restricted_4.c,
* gcc.target/aarch64/sve/acle/general-c/load_gather_sv_restricted_1.c,
* gcc.target/aarch64/sve/acle/general-c/load_gather_sv_restricted_2.c,
* gcc.target/aarch64/sve/acle/general-c/mul_lane_1.c,
* gcc.target/aarch64/sve/acle/general-c/shift_left_imm_long_1.c,
* gcc.target/aarch64/sve/acle/general-c/shift_left_imm_to_uint_1.c,
* gcc.target/aarch64/sve/acle/general-c/shift_left_imm_to_uint_2.c,
* gcc.target/aarch64/sve/acle/general-c/shift_right_imm_narrowb_1.c,
* gcc.target/aarch64/sve/acle/general-c/shift_right_imm_narrowb_to_uint_1.c,
* gcc.target/aarch64/sve/acle/general-c/shift_right_imm_narrowt_1.c,
* gcc.target/aarch64/sve/acle/general-c/shift_right_imm_narrowt_to_uint_1.c,
* gcc.target/aarch64/sve/acle/general-c/store_scatter_index_restricted_1.c,
* gcc.target/aarch64/sve/acle/general-c/store_scatter_offset_restricted_1.c,
* gcc.target/aarch64/sve/acle/general-c/tbl_tuple_1.c,
* gcc.target/aarch64/sve/acle/general-c/ternary_long_lane_1.c,
* gcc.target/aarch64/sve/acle/general-c/ternary_long_opt_n_1.c,
* gcc.target/aarch64/sve/acle/general-c/ternary_qq_lane_rotate_1.c,
* gcc.target/aarch64/sve/acle/general-c/ternary_qq_rotate_1.c,
* gcc.target/aarch64/sve/acle/general-c/ternary_shift_right_imm_1.c,
* gcc.target/aarch64/sve/acle/general-c/ternary_uint_1.c,
* gcc.target/aarch64/sve/acle/general-c/unary_convert_narrowt_1.c,
* gcc.target/aarch64/sve/acle/general-c/unary_narrowb_1.c,
* gcc.target/aarch64/sve/acle/general-c/unary_narrowb_to_uint_1.c,
* gcc.target/aarch64/sve/acle/general-c/unary_narrowt_1.c,
* gcc.target/aarch64/sve/acle/general-c/unary_narrowt_to_uint_1.c,
* gcc.target/aarch64/sve/acle/general-c/unary_to_int_1.c: New tests.
* gcc.target/aarch64/sve2/bcax_1.c: Likewise.
* gcc.target/aarch64/sve2/acle: New directory.

From-SVN: r280060

[AArch64] Pass a mode to some SVE immediate queries

It helps the SVE2 ACLE support if aarch64_sve_arith_immediate_p and
aarch64_sve_sqadd_sqsub_immediate_p accept scalars as well as vectors.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
(aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
* config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
(aarch64_sve_sqadd_sqsub_immediate_p): Likewise.  Handle scalar
immediates as well as vector ones.
* config/aarch64/predicates.md (aarch64_sve_arith_immediate)
(aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
(aarch64_sve_qsub_immediate): Update calls accordingly.

From-SVN: r280059

[AArch64] Add banner comments to aarch64-sve2.md

This patch imposes the same sort of structure on aarch64-sve2.md
as we already have for aarch64-sve.md, before it grows a lot more
patterns.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/aarch64-sve2.md: Add banner comments.
(<su>mulh<r>s<mode>3): Move further up file.
(<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
(*aarch64_sve2_sra<mode>): Move further down file.
* config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.

From-SVN: r280058

compiler: don't localize names in export data

    Localizing names in export data causes the compiler output to change
    depending on the LANG environment variable, so don't do it.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/214038

From-SVN: r280057

compiler: don't add composite literal keys to package bindings

    Adding composite literal keys to package bindings gets confusing when
    it is combined with dot imports.  The test case showing the resulting
    compilation failure is https://golang.org/cl/213899.

    Fix this by adding a new expression type to hold composite literal keys.
    We shouldn't see it during lowering if it is a struct field name,
    because Composite_literal_expression::do_traverse skips struct field names.
    Or, it should, but that didn't quite work with pointer types so it had to
    be tweaked.

    This lets us remove the code that recorded whether an Unknown_expression
    is a composite literal key.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/214017

From-SVN: r280056

[amdgcn] Add support for sub-word sync_compare_and_swap operations

2020-01-09  Kwok Cheung Yeung  <kcy@codesourcery.com>

libgcc/
* config/gcn/atomic.c: New.
* config/gcn/t-amdgcn (LIB2ADD): Add atomic.c.

From-SVN: r280055

[AArch64] Simplify WHILERW and WHILEWR definition

I'd made WHILERW and WHILEWR use separate patterns from the SVE
WHILE instructions, but they're similar enough that we can use
a single pattern.  This means that we also get the flag-related
patterns "for free".

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
and UNSPEC_WHILEWR.
(while_optab_cmp): Handle them.
* config/aarch64/aarch64-sve.md
(*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
and add a "@" marker.
* config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
instead of gen_aarch64_sve2_while_ptest.
(@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.

From-SVN: r280054

[AArch64] Rename UNSPEC_WHILE* to match instruction mnemonics

The UNSPEC_WHILE*s had an underscore before the condition code,
whereas almost all other SVE unspecs are taken directly from
the mnemonic.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
(UNSPEC_WHILELE): ...this.
(UNSPEC_WHILE_LO): Rename to...
(UNSPEC_WHILELO): ...this.
(UNSPEC_WHILE_LS): Rename to...
(UNSPEC_WHILELS): ...this.
(UNSPEC_WHILE_LT): Rename to...
(UNSPEC_WHILELT): ...this.
* config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
(cmp_op, while_optab_cmp): Likewise.
* config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
* config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
(svwhilelt): Likewise.

From-SVN: r280053

[AArch64] Rename SVE shape "unary_count" to "unary_to_uint"

The SVE ACLE shape names use "_int" and "_uint" for arguments that are
signed-integer or unsigned-integer variants of the main vector type.
With SVE2 this variation also becomes common for return values,
which the main SVE2 patch handles using "_to_int" and "_to_uint".
This patch renames the existing unary_count shape to match the
new scheme.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
(unary_to_uint): Define.
* config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
(unary_count): Rename to...
(unary_to_uint_def, unary_to_uint): ...this.
* config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.

gcc/testsuite/
* gcc.target/aarch64/sve/acle/general-c/unary_count_1.c: Rename to...
* gcc.target/aarch64/sve/acle/general-c/unary_to_uint_1.c: ...this.
* gcc.target/aarch64/sve/acle/general-c/unary_count_2.c: Rename to...
* gcc.target/aarch64/sve/acle/general-c/unary_to_uint_2.c: ...this.
* gcc.target/aarch64/sve/acle/general-c/unary_count_3.c: Rename to...
* gcc.target/aarch64/sve/acle/general-c/unary_to_uint_3.c: ...this.

From-SVN: r280052

[AArch64] Specify some SVE ACLE functions in a more generic way

This patch generalises some boilerplate that becomes much more
common with SVE2 intrinsics.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/aarch64-sve-builtins-functions.h
(code_for_mode_function): New class.
(CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
* config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
(svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
(svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
(svmul_lane, svtmad): Use CODE_FOR_MODE0.

From-SVN: r280051

[AArch64] Tweak iterator usage for [SU]Q{ADD,SUB}

The pattern:

;; <su>q<addsub>

(define_insn "aarch64_<su_optab><optab><mode>"
  [(set (match_operand:VSDQ_I 0 "register_operand" "=w")
(BINQOPS:VSDQ_I (match_operand:VSDQ_I 1 "register_operand" "w")
  (match_operand:VSDQ_I 2 "register_operand" "w")))]
  "TARGET_SIMD"
  "<su_optab><optab>\\t%<v>0<Vmtype>, %<v>1<Vmtype>, %<v>2<Vmtype>"
  [(set_attr "type" "neon_<optab><q>")]
)

meant that we overloaded "optab" to be "qadd" for both SQADD and
UQADD.  Most other "optab" entries are instead the full optab name,
which for these patterns would be "ssadd" and "usadd" respectively.
(Unfortunately, the optabs don't extend to vectors yet, something
that would be good to fix in GCC 11.)

This patch therefore does what the comment implies and uses
q<addsub> to distinguish qadd and qsub instead.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/iterators.md (addsub): New code attribute.
* config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
Re-express as...
(aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
in the asm string and attributes.  Fix indentation.
* config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
Re-express as...
(@aarch64_sve_<optab><mode>): ...this.
* config/aarch64/aarch64-sve-builtins.h
(function_expander::expand_signed_unpred_op): Delete.
* config/aarch64/aarch64-sve-builtins.cc
(function_expander::expand_signed_unpred_op): Likewise.
(function_expander::map_to_rtx_codes): If the optab isn't defined,
try using code_for_aarch64_sve instead.
* config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
(svqsub_impl): Likewise.
(svqadd, svqsub): Use rtx_code_function instead.

From-SVN: r280050

[AArch64] Remove fictitious [SU]RHSUB instructions

We've had skeleton support for "SRHSUB" and "URHSUB" since the initial
commit of the port, but no such instructions exist.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
(HADDSUB, sur, addsub): Remove them.

From-SVN: r280049

Fix tree-nrv.c ICE for direct internal functions

pass_return_slot::execute has:

      /* Ignore internal functions without direct optabs,
 those are expanded specially and aggregate_value_p
 on their result might result in undesirable warnings
 with some backends.  */
      && (!gimple_call_internal_p (stmt)
  || direct_internal_fn_p (gimple_call_internal_fn (stmt)))
      && aggregate_value_p (TREE_TYPE (gimple_call_lhs (stmt)),
    gimple_call_fndecl (stmt)))

But what the comment says applies to directly-mapped internal functions
too, since they're only used if the target supports them without a
libcall.

This was triggering an ICE on the attached testcase.  The svld3 call
is folded to an IFN_LOAD_LANES, which returns an array of vectors with
VNx48QImode.  Since no such return type can exist in C, the target hook
was complaining about an unexpected use of SVE modes.  (And we want to
keep asserting for that, so that we don't accidentally define an ABI for
an unexpected corner case.)

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* tree-nrv.c (pass_return_slot::execute): Handle all internal
functions the same way, rather than singling out those that
aren't mapped directly to optabs.

gcc/testsuite/
* gcc.target/aarch64/sve/acle/general/nrv_1.c: New test.

From-SVN: r280048

Add a compatible_vector_types_p target hook

One problem with adding an N-bit vector extension to an existing
architecture is to decide how N-bit vectors should be passed to
functions and returned from functions.  Allowing all N-bit vector
types to be passed in registers breaks backwards compatibility,
since N-bit vectors could be used (and emulated) before the vector
extension was added.  But always passing N-bit vectors on the
stack would be inefficient for things like vector libm functions.

For SVE we took the compromise position of predefining new SVE vector
types that are distinct from all existing vector types, including
GNU-style vectors.  The new types are passed and returned in an
efficient way while existing vector types are passed and returned
in the traditional way.  In the right circumstances, the two types
are inter-convertible.

The SVE types are created using:

      vectype = build_distinct_type_copy (vectype);
      SET_TYPE_STRUCTURAL_EQUALITY (vectype);
      TYPE_ARTIFICIAL (vectype) = 1;

The C frontend maintains this distinction, using VIEW_CONVERT_EXPR
to convert from one type to the other.  However, the distinction can
be lost during gimple, which treats two vector types with the same
mode, number of elements, and element type as equivalent.  And for
most targets that's the right thing to do.

This patch therefore adds a hook that lets the target choose
whether such vector types are indeed equivalent.

Note that the new tests fail for -mabi=ilp32 in the same way as other
ACLE-based tests.  I'm still planning to fix that as a follow-on.

2020-01-09  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
* target.def (compatible_vector_types_p): New target hook.
* hooks.h (hook_bool_const_tree_const_tree_true): Declare.
* hooks.c (hook_bool_const_tree_const_tree_true): New function.
* doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
* doc/tm.texi: Regenerate.
* gimple-expr.c: Include target.h.
(useless_type_conversion_p): Use targetm.compatible_vector_types_p.
* config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
function.
(TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
* config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
Use the original predicate if it already has a suitable type.

gcc/testsuite/
* gcc.target/aarch64/sve/pcs/gnu_vectors_1.c: New test.
* gcc.target/aarch64/sve/pcs/gnu_vectors_2.c: Likewise.

From-SVN: r280047

Fortran] PR84135 fix merging dimension into codimension array spec

        PR fortran/84135
        * array.c (gfc_set_array_spec): Fix shifting of codimensions
        when adding a dimension.
        * decl.c (merge_array_spec): Ditto. Fix using correct codimensions.

        PR fortran/84135
        * gfortran.dg/coarray/codimension_3.f90: New.

From-SVN: r280046

libstdc++: Define memory resource key functions non-inline (PR93208)

This prevents the vtables and RTTI from being emitted in every object
file that uses memory_resource and monotonic_buffer_resource.

Objects compiled by GCC 9.1 or 9.2 will contain inline definitions of
the destructors, vtable and RTTI, but this is harmless. The inline
definitions have identical effects to the ones that are now defined in
libstdc++.so so it doesn't matter if the inline ones are used instead of
calling the symbols exported from the runtime library.

PR libstdc++/93208
* config/abi/pre/gnu.ver: Add new exports.
* include/std/memory_resource (memory_resource::~memory_resource()):
Do not define inline.
(monotonic_buffer_resource::~monotonic_buffer_resource()): Likewise.
* src/c++17/memory_resource.cc (memory_resource::~memory_resource()):
Define.
(monotonic_buffer_resource::~monotonic_buffer_resource()): Define.
* testsuite/20_util/monotonic_buffer_resource/93208.cc: New test.

From-SVN: r280044

Make cgraph_edge::resolve-speculation static

2020-01-09  Martin Jambor  <mjambor@suse.cz>

* cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
resolve_speculation and redirect_call_stmt_to_callee static.  Change
return type of set_call_stmt to cgraph_edge *.
* auto-profile.c (afdo_indirect_call): Adjust call to
redirect_call_stmt_to_callee.
* cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
make the this pointer explicit, adjust self-recursive calls and the
call top make_direct.  Return the resulting edge.
(cgraph_edge::remove): Make this pointer explicit.
(cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
(cgraph_edge::make_direct): Likewise, adjust call to
resolve_speculation.
(cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
call to set_call_stmt.
(cgraph_update_edges_for_call_stmt_node): Update call to
set_call_stmt and remove.
* cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
Renamed edge to master_edge.  Adjusted calls to set_call_stmt.
(cgraph_node::create_edge_including_clones): Moved "first" definition
of edge to the block where it was used.  Adjusted calls to
set_call_stmt.
(cgraph_node::remove_symbol_and_inline_clones): Adjust call to
cgraph_edge::remove.
* cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
make_direct and redirect_call_stmt_to_callee.
* ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
resolve_speculation and make_direct.
* ipa-inline-transform.c (inline_transform): Adjust call to
redirect_call_stmt_to_callee.
(check_speculations_1):: Adjust call to resolve_speculation.
* ipa-inline.c (resolve_noninline_speculation): Adjust call to
resolve-speculation.
(inline_small_functions): Adjust call to resolve_speculation.
(ipa_inline): Likewise.
* ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
make_direct.
* ipa-visibility.c (function_and_variable_visibility): Make iteration
safe with regards to edge removal, adjust calls to
redirect_call_stmt_to_callee.
* ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
and redirect_call_stmt_to_callee.
* multiple_target.c (create_dispatcher_calls): Adjust call to
redirect_call_stmt_to_callee
(redirect_to_specific_clone): Likewise.
* tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
Adjust calls to cgraph_edge::remove.
* tree-inline.c (copy_bb): Adjust call to set_call_stmt.
(redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
(expand_call_inline): Adjust call to cgraph_edge::remove.

From-SVN: r280043